KR20100113503A - System and method for storing program code and data within an application hosting center - Google Patents

Abstract

Systems and methods for storing program code and data in an application hosting center are disclosed. For example, a computer-implemented method of one embodiment may include partitioning program code and / or data used to execute an online application into a first type and a second type; Storing the first type of program code and data in a first type of memory that provides relatively low latency memory access; Storing the second type of program code and data in a second type of memory that provides relatively higher latency memory access than the first type of memory; Retrieving program code and data from the first and second types of memory in response to a client request to execute an online application; And transmitting to the client a streaming interactive video stream representing an image generated by the application.

Description

SYSTEM AND METHOD FOR STORING PROGRAM CODE AND DATA IN AN APPLICATION HOSTING CENTER

This application is a continuation-in-part (CIP) filed 10 / 315,460 filed Dec. 10, 2002 entitled “Method and Apparatus for Wireless Video Gaming,” by the assignee of the CIP application. Was positive.

The present invention generally relates to the field of data processing systems that enhances a user's ability to access and manipulate audio and video media.

Recorded audio and video media have been an aspect of society since the day of Thomas Edison. At the beginning of the twentieth century, recorded audio media (cylinders and records) and video media (nickelodeons and movies) became widespread, but both technologies were still in their infancy. In the late 1920s, in the mass market, movies were combined with audio, followed by color video with audio. Radio broadcasting has evolved significantly in the form of ad support for broadcast mass market audio media. When television broadcast standards were established in the mid-1940s, television combined with radio as a form of broadcast mass-market medium that already recorded or brought live movies home.

By the mid-twentieth century, the majority of American households used a phonograph record player to play recorded audio media, radio to receive live audio, and live broadcast audio / video (A / V) media. I had a television set. Very often these three "media players" (record players, radios and televisions) are combined into a cabinet that shares a common speaker that becomes a "media center" in the home. Although media choice was limited to consumers, the media "ecosystem" was quite stable. Most consumers knew how to use "media players" and enjoyed the full range of their capabilities. At the same time, media publishers (primarily video and television studios and record companies) do not suffer from "second sales" or extensive piracy, such as resale of used media. Could distribute to theaters and homes. A typical publisher does not profit from the second sale, which alone reduces the revenue from buyers of used media for new sales. Although used records may have been sold in the mid-20th century, such sales had little impact on record publishers. Because unlike a movie or video program—one or a few times watched by adults—the music tracks will be heard hundreds or thousands of times. So music media is much less "perishable" (e.g. of lasting value to adult consumers) than movie / video media. Once a record is purchased, if the consumer likes music, the consumer will be able to hold it for a long time.

From the mid-20th century to the present day, the media environment has undergone a series of radical changes in the interests and losses of consumers and publishers. There was certainly a high degree of consumer convenience, according to the widespread use of audio records, especially cassette tapes with high quality stereo sound. But it also shows that consumer media is the beginning of widespread enforcement. Certainly, many consumers have used cassette tapes that purely taped their own records for convenience, but more and more consumers (eg, students in dormitories with easy access to each other's record collections) will make copyrighted copies . In addition, the consumer will tape record music played over the radio rather than buying a record or tape from the publisher.

The advent of consumer VCRs has brought more consumer convenience, and since the VCR is now installed to record TV shows that can be watched late, it also brought about the creation of a video rental business, as well as TV programming as well as movies. On demand ". Since the mid-1980s, the rapid development of the mass market of home media devices has led to an unprecedented level of convenience and choice for consumers, and to a rapid expansion of the media publishing market.

Today, most consumers are constrained by a particular type of media or a particular publisher, facing an excess of media choices as well as an excess of media devices. An avid consumer of media owns a number of devices connected to televisions and computers in various rooms of the home, so that not only a group of wireless controls, but also a "rat's nest" of personal computers (PCs) and / or one or more televisions and cables ) ". (In the context of this application, "Personal Computer" or "PC" refers to any kind of computer suitable for home or office, which is a desktop, Macintosh ^® or other non-Windows-based, Windows-based device). These devices include video game consoles, VCRs, DVD players, audio sorround-sound processors / amplifiers, and satellite set-tops. Boxes (satellite set-top box), cable TV set-top box, and the like. And for the frenzied consumer, there are a number of similar functions because of compatible issues. For example, a consumer would own HD-DVD and Blu-ray DVD players, or Microsoft XBOX ^® , Sony Playstation ^® video game systems. Moreover, because of the incompatibility of game console versions between certain games, consumers will own both later versions, such as XBox and XBox 360 ^® . Often, consumers are bewildered about which video input and which remote operation to use. Even after the disc is located in the correct player (e.g. DVD, HD-DVD, Blu-ray, Xbox or Playstation), its video and audio inputs are selected for the device, and the correct remote control is found, the consumer still remains a technical challenge Will face. For example, in the case of widescreen DVDs, the user needs to first determine and set the correct aspect ratio (eg 4: 3, Full, Zoom, Wide Zoom, Cinema Wide, etc.) for the TV or monitor screen. It becomes Similarly, the user needs to first determine and then set the correct audio surround sound system format (e.g., AC-3, Dolby Digital, DTS, etc.). Every time, consumers don't know that they don't enjoy the media content with the full performance of their television or audio system (e.g. watching a movie distorted with the wrong aspect ratio, or audio in stereo rather than surround sound). Listening).

Increasingly, Internet-based media devices are added to a stack of devices. Audio devices such as the Sonos ^® Digital Music System stream audio directly from the Internet. Similarly, devices such as the Slingbox ^™ entertainment player can stream, record, or watch video away from the PC via the home network. In addition, Internet Protocol Television (IPTV) services often provide services such as cable TV through digital subscriber line (DSL) or other home Internet connections. There is also recent effort to integrate multiple media functions into a single device, such as PCs running Moxi ^® Media Center and Windows XP Media Center Edition. While each of these devices provides an element of convenience for the function that performs them, each lacks ubiquitous and simple access to most media. Moreover, due to frequent expensive processing and / or the need for local storage, such devices are costing hundreds of dollars for production. In addition, these modern consumer electronic devices typically consume a lot of power even when they are not in operation, which means that they waste energy resources and become expensive over time. For example, some devices will continue to operate if the consumer ignores turning them off or switching to another video input. And since no device is a complete solution, it must be integrated with a stack of other devices in the home, which still leaves the user on the sea and chaotic wires of the remote control.

Moreover, when many new Internet-based devices work properly, they typically offer media in a generic form than is available in other forms. For example, a device that streams video over the Internet is a video material other than the interactive "extras" that often accompany DVDs, such as "making of" video, games, or director's commentary. stream the material. This is often due to the fact that the interactive material is made in a particular format intended for a particular device that handles interactivity locally. For example, DVD, HD-DVD, and Blu-ray Discs each have their own specific interactive format. Some home media devices or local computers that have been developed to support all popular formats require a level of sophistication and flexibility that is complex and fairly expensive for consumers to operate.

In addition to that problem, if a new format is introduced in the future, the local device will not have the hardware capability to support the new format, which means that the consumer must purchase an upgraded local media device. For example, if higher-resolution video or stereoscopic video (eg, one video stream for each eye) has been introduced recently, the local device may not have computer capability to decode the video, or In new formats (e.g., if stereophonic sound is achieved through 120 fps video synchronized with shuttered glasses delivering 60 fps to each eye, this option will not be available in the absence of upgraded hardware). You will not have the hardware to output the video.

The issue of media device obsolescence and complexity is a serious problem when it becomes sophisticated interactive media, especially in video games.

Current video game applications include four major non-portable hardware platforms: Sony PlayStation® 1,2 and 3 (PS1, PS2, and PS3); Microsoft Xbox®; And Nintendo Gamecube® and Wii ^™ ; And PC based games. Each of these platforms is different, so recorded games that run on one platform usually do not run on another. There is also a problem of compatibility from one generation of device to the next. Although the majority of software game developers create software that is designed to be platform-independent, a proprietary layer of software (often referred to as the "game development engine") to run a particular game on a particular platform. Is required to adapt the game for use on a particular platform. Each platform is sold to the consumer as a "console" (eg, standalone box attached to a TV or monitor / speaker) or it is the PC itself. Typically, video games are sold on optical media such as Blu-ray, DVD, DVD-ROM or CD-ROM, which include video games embodied as sophisticated real-time software applications.

Particular needs to achieve a platform compatible with video game software will be inevitable due to real-time characteristics and the high computational requirements of advanced video games. For example, from one generation of video games to the next generation (e.g., Xbox to Xbox 360 or Playstation 2), as is the general compatibility of productivity applications (e.g. Misrosoft Word) with other faster processing units and cores on one PC. One can expect a game that is sufficiently compatible with Playstation 3 ("PS3")). However, this is not the case for video games. Because typically video game manufacturers look for the best possible performance at a given price point when a video game generation is released, and breakthrough structural changes in the system frequently occur so that many games recorded for previous generation systems are the latest generation. It will not work on the system. For example, XBox is based on the x86-series processor, while XBox 360 is based on the PowerPC family.

While techniques for emulating previous architectures can be used, a given video game is a real-time application and cannot be executed to achieve exactly the same behavior in emulation. This is a loss for consumers, video game console manufacturers, and video game software publishers. For consumers, this means keeping both old and new generations of video game consoles connected to TVs to play all games. That means the costs associated with later adoption and emulation of new consoles for console manufacturers. And for publishers, different versions of new games must be released to reach all potential consumers--not only for each brand of video game (e.g., XBox, Playstation), but also for existing brands (e.g. PS2 and PS3). That means releasing a version for each version. For example, Electronic Art's "Madden NFL 08" was developed for XBox, XBox360, PS2, PS3, Gamecube, Wii, and PC, and other platforms.

Portable devices such as cellular (“cell”) phones and portable media players are also challenging game developers. Increasingly, such devices can connect to wireless data networks and download video games. However, there are a wide variety of cell phones and media devices with different display resolutions and computing capabilities in the market. Also, as such devices typically have power consumption, price, and weight constraints, they typically have advanced graphics processing units ("GPUs"), which are devices made by NVIDIA, Santa Clara, CA. Lack of the same graphics acceleration hardware. As a result, game software developers typically develop existing game titles for many different types of portable devices at the same time. One user will find that an existing game title cannot be used on his cell phone or portable media player.

In the case of home game consoles, hardware platform manufacturers typically charge royalties to software game developers to release games on their platforms. Cell phone wireless carriers typically charge royalties to game publishers for downloading games to cell phones. In the case of PC games, there are no royalties paid for published games, but game developers typically pay a high cost due to the high customer service and possible installation issues they support to support a wide range of PC configurations. Face problems. In addition, PCs are typically easily reprogrammed by technically knowledgeable users, so the barrier to piracy of game software is low, and games are more easily looted and distributed (eg, via the Internet). Thus, for software game developers, there are disadvantages and costs to launching on game consoles, cell phones and PCs.

For console and PC software game publishers, the cost is not the end there. To distribute games through retail channels, the publisher charges a wholesale price below the selling price so that the retailer can make a profit. Publishers also typically have to pay for the production and dissemination of the physical media containing the game. Publishers may also choose to be "sponsored" by the retailer to cover possible contingencies, such as if the game does not sell, the price of the game falls, or if the retailer refunds some or all of the wholesale price and / or exchanges the game from the buyer. You will be charged a price protection fee. In addition, the retailer also typically charges the publisher a fee to help sell the game in the advertising leaflet. Moreover, retailers gradually buy back games from users who have played, then sell them as used games, and typically distribute none of the game publishers and used game revenue. The added cost to game publishers is that games are often pirated and distributed over the Internet, which users download and copy for free.

Internet broadband speeds have been increasing and broadband connectivity has become more widespread in the US and around the world, especially with home-connected PCs on the Internet and cafés, and games are increasingly being downloaded for PCs or consoles. It is distributed through. In addition, broadband connections are increasingly used to play multiplayer and mass multiuser online games (both referred to herein as "MMOG"). These changes alleviate the costs and issues associated with retail distribution. Downloading online games presents some disadvantages to game publishers in that distribution costs are typically low and little or no cost from unsold media. However, downloaded games are still susceptible to copyright infringement, and because of their size (often many gigabytes in size), they can take a lot of time to download. In addition, multiple games can fill small disk drives, such as those sold with portable computers or with video game consoles. However, extended games or MMOGs require an online connection with the game to play, and the copyright infringement problem can be mitigated by requiring that the user typically have a valid user account. Unlike linear media (eg, video and music) that can be replicated by a microphone recording audio from a speaker or a camera taking video on a display screen, each video game experience is unique and simply video Can't be duplicated using audio recording. Thus, even in areas where copyright law exists, illegal sanctions are not available, and piracy is rampant. MMOGs can be protected from piracy and so businesses can be supported. For example, Vivendi SA's "World of Warcraft" MMOG has been successfully deployed without piracy worldwide. And many online or MMOG games, such as Linden Lab's "Second Life" MMOG, generate revenue for the operator of the game through an economic model built on games where capital can be created using buy, sell and online tools. Create. Thus, mechanisms added to a typical game software purchase or subscription can be paid for use of the online game.

Piracy, on the other hand, can often be mitigated due to the nature of online or MMOGs, and online game operators are still faced with remaining challenges. Many games require substantially local (eg, home) processing resources for online or MMOG to work properly. If the user has a low-performance local computer (eg, no GPU, such as a low-performance laptop), he cannot play the game. Moreover, depending on the age of game consoles, they are far behind the latest technology and will no longer be able to handle advanced games. Although the user's local PC can handle the computer's computer requirements, there is often a complex setup. There may be incompatibilities in the driver (eg, if a new game is downloaded, a previously installed game that depends on the older version of the graphics driver will install a new version of the graphics driver that will not work). The console will run out of local disk space as many games are downloaded. Complicated games can be created over time as bugs are found and fixed, or if changes are made to the game (eg, if the game developer finds the level of the game too difficult or too easy to play). Thus typically receive the downloaded patch. These patches require a new download. However, sometimes not all users complete the download of all patches. On the other hand, downloaded patches introduce issues such as other compatibility and disk space consumption.

In addition, while playing a game, many data downloads are required to provide graphical or operational information to the local PC or console. For example, if a user enters an MMOG room and encounters a scene or character consisting of graphical data and actions that are not available to the user's local machine, the location or character's data must be downloaded. This causes a substantial delay during game play if the internet connection is not fast enough. And if the places and characters faced demand beyond the computer's capabilities of the storage space or the local PC or console, this results in a situation where the user cannot proceed in the game or continue with low-quality graphics. Thus, online or MMOG games often limit their storage and / or computer complex needs. Moreover, it often limits the amount of data transfer during the game. Online or MMOG games may also narrow the market for users who can play the game.

In addition, a technically knowledgeable user will gradually reverse-engineer local copies of the game and can deceive them by transforming the game. Cheats will be as easy as making a button press faster than humans can (eg, shooting very repeatedly). In games that support in-game asset transactions, the trick can reach a sophisticated level that results in fraudulent transactions involving assets of real economic value. When an online or MMOG economic model is based on such capital transactions, this can lead to substantial detrimental effects for game operators.

Development costs for new games are increased by allowing PCs and consoles to produce increasingly sophisticated games (eg, more realistic graphics such as real-time ray-tracing, and more realistic behaviors such as real-time physical simulation). . In the early days of the video game business, video game development was very similar to the process for developing application software, i.e., most of the development costs were developed for motion pictures with a wide range of special effects, such as graphics, audio, and motion. Contrary to traditional elements or "assets", software development has begun. Today, many sophisticated video game developments try to more closely resemble rich special effect motion picture development than software development. For example, many video games provide simulations of 3-D worlds, and gradually create photorealistic (eg, photorealistic computer graphics like photorealistic images taken with photographics) characters, props, and environments. One of the most challenging aspects of photorealistic game development is the creation of computer-generated human faces that are indistinguishable from living human faces. Face captured, such as the Contour ^TM Reality Capture developed by Mova of San Francisco, California, technology, track, and capture the face of a running motion with the correct geometry for high resolution. This will allow 3D faces to be rendered on a PC or game console so that they are not visually indistinguishable from captured live facial movements. Capturing and rendering the human face exactly as a "photoreal" is useful in many respects. First, recognizable celebrities or athletes are often used in video games (often employed at high cost), and defects can be visible to the user, creating an unfocused or unpleasant observation experience. Often, a high degree of detail is potentially a face movie as a face movie, which requires the rendering of multiple polygons and / or high resolution textures with polygons and / or textures that vary from frame to frame. It is required to achieve realism.

When high polygon-count scenes with detailed textures change rapidly, the PC or game console that supports the game has enough space to store enough polygon and texture data for the required number of animation frames created in the game segment. You will not have RAM. Furthermore, single optical drivers or single disk drivers typically used in PCs or game consoles are usually slower than RAM, and typically the maximum transfer rate (i.e. data rate) that the GPU can allow to render polygons and textures. rate)) may not be maintained. Current games typically load most polygons and textures into RAM, which means that a given scene is usually limited in complexity and persistence by the performance of the RAM. In the case of facial animation, for example, it is limited to either a low resolution face that is not a photorealistic picture, or a photorealistic face that can only be animated with a limited number of frames before the game or game console stops. Load for many frames.

Watching a progress bar move slowly across the screen as a PC or console displays a message similar to "loading ..." is an inherent drawback by today's users of complex video games. The next scene is loaded from a disc (here "disk" is referred to as non-volatile optical or magnetic media as well as non-disk media such as semiconductor "flash" memory, unless otherwise limited). The delay can take several seconds or minutes. This is a waste of time and can be quite embarrassing for game players. As discussed above, much or all of the delay will be due to the load time for data from polygons, textures or other disks, but may also be spent while the GPU on the processor and / or PC or console prepares the data for the scene. It will be part of the load time. For example, football video games allow players to select many players, teams, playgrounds, weather conditions, and the like. Thus, depending on the particular combination selected, different polygons, textures and other data (collectively "object") are required for the scene (e.g., different teams may assign different colors and patterns to their uniforms). Have). It would be possible to store objects on the disk used to store the game, precompute all objects or many objects in advance, and enumerate many or all the various permutations. But if the substitution is too large, much of the storage required for all objects will be too large (or not too practical to download) to the disk. Thus, existing PC or console systems are typically forced to both complexity and duration of a given scene, resulting in long load times for complex scenes.

Another significant limitation of prior art video game systems and application software systems is the gradual use of large databases of 3D objects, such as polygons and textures, which need to be loaded into a PC or game console for processing. As mentioned above, such a database took a long time to load when stored locally on disk. However, load time is even more serious if the database is stored at a remote location and can be accessed over the Internet. In such a situation, it may take several minutes, hours, or days to download many databases. Even more, such databases are often made very expensive (eg 3D models of detailed tall-masted sailboats used in games, movies, or historical documentaries) and are intended for sale to local end-users. . However, there is a risk of piracy as the database is downloaded to local users. In many cases, the user simply selects a database for evaluation reasons to see if it suits the user's needs (e.g., the 3D costume for the game character has a satisfactory appearance or appearance when the user performs a particular movement). I want to download it. Long load times can restrain the user from evaluating the 3D database before making a purchase decision.

Similar issues arise, especially in MMOGs such as games that allow users to gradually use custom characters, such as games. For a PC or game console to display a character, not only the behavior for the character (eg, if the character has a shield, is the shield strong enough to avoid spears), and the 3D geometry database You need access to (polygons, textures, etc.). Typically, when an MMOG is first played by a user, many databases for the character can already be used as an initial copy of the game, which is downloaded to the game's optical disc or disk and used locally. However, as the game progresses, if a user encounters a database of characters or objects, their database is not available locally (eg, if another user has created a custom character) before the character or object is displayed. , Its database must be downloaded. This substantially causes game delays.

The sophistication and complexity of a given video game, another challenge for video game developers and publishers with prior art video game consoles, will often take two to three years at a cost of tens of millions of dollars to develop a video game. Given that new video game console platforms are introduced at a rate of approximately every five years, game developers will be able to develop these game development tasks in advance of the launch of new game consoles in advance of the launch of new game consoles to have video games available at the same time when the new platform is released. It is necessary to start. Some consoles from competitors are sometimes released at the same time (eg within a year or two), or the remainder is the popularity of the console, which will make the console the largest video game software sales. For example, in the recent console cycle, Microsoft XBox 360, Sony Playstation 3, and Nintendo Wii were introduced in the same general period, but in the years before they were introduced, game developers were asked about whether the console platform would be more successful than others. "Place their bets" and commit to matching their development resources. Video production companies should also allocate their limited production resources based on evaluating whether the film will be very successful prior to the release of the film. Given the increasing level of investment required for video games, game production will increase gradually as video production, and game production companies will typically use their production resources based on their assessment of the future success of a particular video game. However, unlike video production companies, this conjecture It is not based solely on the success of the production itself; quite, it is predicted based on the success of the game console that the game is intended to run in. At the same time releasing the game on multiple consoles mitigates the risk, but this additional effort is costly. Increase and often delay the actual release of the game.

Application software and user environments on PCs are more computer intensive, dynamic and interactive, not only making them more visually visible to users, but also making them more useful and easier to use. For example, successive versions of the new Window Vista ^™ operating system and the Macintosh® operating system include visual animation effects. Advanced graphics tools such as Maya ^{TM from} Autodesk, Inc. offer sophisticated 3D rendering and animation capabilities that limit the very latest CPUs and GPUs. However, the computer requirements of these new tools create many practical issues for users and software developers of such products.

Since the visual display of the operating system (OS) is no longer sold, but must still operate on a wide range of computers, including older generations of computers that can be upgraded to the new OS, OS graphics requirements Is limited to a large extent by the minimum common requirements of the target computer, which typically includes a computer without a GPU. This severely limits the graphics performance of the OS. Moreover, portable computer (eg, laptop) battery-powered limits visual display performance because high computational performance in a CPU or GPU typically results in high power consumption and short battery life. Portable computers include software that automatically lowers processor activity to reduce power consumption when a process is not used. In some computer models, the user will manually lower the operation of the processor. For example, Sony's VGN-SZ280P laptops have "stamina" on one side (longer battery life for lower performance) and "Speed" on the other side (shorter battery life for higher performance). And a switch labeled. Running an OS on a portable computer should allow the computer to function even when it is operating at some of its peak performance. Thus, OS graphics performance often lags far below the available computer capabilities of the state of the art.

High-end, computer-intensive applications such as Maya are often sold with the expectation that they will be used in high-performance PCs. This generally establishes at least common denominator requirements of very high performance, more cost and less portability. As a result, such applications have a much limited target audience (or general purpose production applications, such as Microsoft Office) than general purpose OSs, and are typically sold on a much smaller scale than general purpose OS software or general purpose application software. Potential consumers are further limited because it is difficult for future users to attempt such computer intensive applications in advance. For example, consider a student wanting to learn how to use Maya, or a potential buyer who already has knowledge of such an application wants to try to use Maya before investing in a purchase (which is a high-end that can run Maya). Consider buying a computer as well). On the other hand, students or potential buyers may download, but you can get a physical media copy of Maya's demo version, and if they are using Maya as a computer's full potential (eg dealing with complex 3D scenes), If the running computer lacks performance, then they will not be able to fully evaluate the product's information. This substantially limits the consumer for such high-end applications. It will also result in higher selling prices, since development costs are split into fewer purchases than general purpose applications.

In addition, expensive applications can create additional incentives for individuals and companies to make pirated copies of application software. As a result, high-end application software suffers widespread piracy despite the considerable efforts of publishers of such software to mitigate such piracy by various techniques. Still, when using pirated high-end applications, users cannot eliminate the need to invest in expensive state-of-the-art PCs to run pirated copies. So, while they can get the use of a software application as part of the actual retail price, users of pirated software still need to get or buy an expensive PC to fully use the application.

This is the same for users of high-performance piracy video games. Although pirates can get games at only a fraction of the actual price, they still ask for expensive computing hardware (such as PCs with enhanced GPUs or high-end video game consoles like Xbox 360) that are needed to play the game properly. do. If video games are generally a hobby for consumers, then additional costs for high-end video game systems may be banned. This situation is exacerbated in countries where the average annual income of workers is now considerably lower than in the United States (eg, China). As a result, a very small percentage of the total population owns high-end video game systems or high-end PCs. In such countries, "Internet cafes", in which users pay for the use of computers connected to the Internet, are quite common. Frequently, these Internet cafes have older models or low-end PCs without high performance features such as GPUs, which allow players to play computer-intensive video games. This is a major factor in the success of games running on low-end PCs, such as Vivendi's "World of Warcraft", which is very successful, and is commonly played in Internet cafes in China. On the other hand, computer-intensive games like "Second life" aren't very easy to play on PCs installed in Chinese Internet cafes. Such games are virtually unacceptable to users who only have access to low-performance PCs in Internet cafes.

In addition, barriers exist for users considering purchasing a video game, and the user will be likely to attempt a demo version of the game by downloading the demo over the Internet from home. Video game demos are often full versions of games with some features disabled or limited gameplay size. This would involve a lengthy process (perhaps time) of downloading gigabytes of data before the game is run or installed on a PC or console. In the case of a PC, it understands whether a special driver (eg DirectX, or OpenGL) is needed for the game, downloads the correct version, installs it, and then decides if the PC can play the game. Will include. The final step will include determining if the PC has sufficient processing (CPU and GPU) performance, sufficient RAM, and a compatible OS (eg, some games running on Windows XP, but not running Vista). Thus, after a lengthy process of attempting to run a video game demo, the user will be able to understand that the video game demo may not play well, and that a given user's PC preferences are present. More seriously, if the user downloads a new driver to run the demo, this driver version may not be compatible with other games or applications that the user regularly uses on the PC, so installing the demo will not play the game beforehand. It will either enable or disable the application. In addition to these barriers that embarrass the user, they create barriers for video game software publishers and video game developers to sell games.

Another problem that is the result of economic inefficiencies is related to the fact that a given PC or game console is usually designed to accommodate what level of performance requirements for applications and / or games. For example, if they have no GPU at all, some PCs have more or less RAM, a slower or faster CPU, and a slower or faster GPU. Some games or applications bring the full computing power of a given PC or console, while not many games and applications. If a user's game or application selection falls into the lack of the best performance of the local PC or console, the user will waste money on the PC or console for features not available. In the case of a console, the console manufacturer will pay more than needed to subsidize the console cost.

Another problem with the marketing and enjoyment of video games involves allowing the user to view other play games before purchasing the game. Several prior art approaches exist for the recording of video games for later replay. For example, US Pat. No. 5,558,339 teaches the recording of game state information, including game controller operation, during "gameplay" at a video game client computer (owned by the same or another user). This state information can later be used to replay some or all of the game actions on the video game client computer (eg, PC or console). A significant disadvantage of this approach is that in order for a user to watch a recorded match, the user must have a video game client computer capable of playing the game and own a video game application running on the computer, so the gameplay is recorded. Same as when the game was replayed. In addition, video game applications can be written in such a way that there is no possible performance difference between the recorded game and the game being played.

For example, game graphics are typically calculated frame by frame. In many games, game logic is sometimes the next depending on whether one scene is particularly complex, or if there are other delays, such as slow down execution (e.g., other processing on the PC will be executed to take CPU cycles from the game application). It may take shorter or longer than one frame time to compute the graphics for displaying the frame. In such a game, a "threshold" frame may eventually occur that computes in much less than one frame time (so-called less than a CPU clock cycle). When the same scene is recalculated using exactly the same game state information, one frame time (e.g., if the internal CPU bus is out of phase with the external DRAM bus and it leads to a delay in CPU cycle time, Nevertheless, game processing may take fewer CPU clock cycles than other processing, which takes a few milliseconds of CPU time. Thus, when the game is played back, the frame is computed and obtained at two frame times rather than a single frame time. Some actions are based on how often the game computes new frames (eg, when the game samples input from the game controller). While the game is playing, this difference in the time criteria for other actions does not affect game play, but it may result in producing different results in the playback game. For example, if the ballistics of a basketball are reliably computed at 60 fps, the game controller input is sampled based on the computed frame rate, and when the game is recorded, the computed frame rate would be 53 fps, but the game would be replayed. As it is 52 fps, this can make a difference between whether the basketball is blocked from entering the basket, resulting in different results. Thus, using game state to record a video game requires the same game state information to produce exactly the same result, and requires very careful game software design to ensure replay.

Another prior art approach for recording video games is simply to record the video output of a PC or video game system (eg, for a video capture board on a PC or for a VCR or DVD recorder). The video may then be rewound and replayed, or alternatively, it may be a recorded video that is typically uploaded to the Internet after being compressed. The disadvantage of this approach is that when a 3D game sequence is played back, the user is limited to seeing the sequence at one point in time when the sequence was recorded. In other words, the user cannot change the viewpoint of the scene.

Moreover, when a compressed video of a recorded game sequence played on a home PC or game console is made available to other users over the Internet, even though the video is compressed in real time, it uploads the compressed video in real time over the Internet. It would be impossible to do. The reason is that many homes around the world are connected to the Internet with high asymmetric broadband connections (eg, DSL and cable modems typically have much greater downstream bandwidth than upstream bandwidth). Compressed high resolution video sequences often have higher bandwidth than the upstream bandwidth performance of the network and make it impossible to upload in real time. Thus, there will be a significant delay (perhaps minutes or hours) after the game sequence is played and before other users can watch the game on the Internet. Although this delay can be tolerable in some situations (eg, seeing the results of a game player that happened in the previous time), it can either eliminate the ability to see the game in real time (eg, basketball tournaments played by champion players). When the game is live playing, remove the "immediate replay" performance.

Another prior art approach allows viewers with television receivers to watch video games live, but only under the control of the television producers. Some television channels provide video game viewing channels in the United States and other countries, where television viewers can watch some video game users (eg, top-ranked player play in tournaments) in the video game channel. This is achieved by having the video output of the video game system (PC and / or console) reflected in the processing mechanism and video distribution for the television channel. This is not the same as a television channel broadcasting live basketball matches in real time, provided by several cameras reflecting different angles around the basketball court. Television channels can then make use of video / audio processing and effects mechanisms to handle output from various video game systems. For example, a television channel may overlay text on the video from a video game that indicates the status of another player (such as may overlay text during a live basketball game), and the television channel may Multiple recordings (overdub) can be made on the narrator's audio that can be discussed. Additionally, the video game output can be combined with the camera recorded video of the game of the actual flare of the game (eg, showing their emotional response to the game).

The problem with this approach is that such live video feeds must utilize the video distribution and processing mechanisms of television channels in real time in order to get the excitement of live broadcast. However, as discussed above, especially if part of the broadcast includes live video from a camera capturing the actual video of the game player, this becomes impossible when the video game system is operating at home. Also, in the tournament situation, there is a concern that gamers at home can modify and cheat the game as mentioned above. For this reason, such video game broadcasts on television channels are often combined in a common location (eg, a television studio or playground) with a television production equipment capable of accepting video reflected from potential live cameras and multiple video game systems. And a player.

Although such a leading video game TV channel can provide television viewers with a very exciting presentation, for example an experience similar to a live sporting event, a video game player that appears as an "athlete", their form of action and the real world in the video game world. In all of their modes of operation, these video systems are limited to situations where the player is physically very close to others. And, after the television channel is broadcast, each broadcast channel can be seen as only one video stream, which is selected by the television manufacturer. Due to this limitation and the high cost of airtime, production equipment and producers, only top-ranked player matches can be watched in top-tier tournaments.

In addition, existing television channels that broadcast full-screen images of video games to television viewers show only one video game at a time. This severely limits the choice of television viewers. For example, a television viewer may not be interested in watching a game at a given time. Another viewer may only be interested in watching the gameplay of a particular player at a given time that is not characterized by the television channel. In other cases, the viewer may be interested in how a professional player can handle a particular level in the game. Still other viewers may wish to control the view of the video game different from the one chosen by the production team. In short, television viewers will have much preference for watching video games that are not supplied by a particular broadcast in a television network, although several other television channels are available. For all of the above reasons, prior art video game television channels have significant limitations in presenting video games to television viewers.

Other drawbacks of prior art video game systems and application software systems are that they are complex and often suffer from errors, crashes and / or unintended and undesirable behavior (collectively, "bugs"). Although games and applications typically undergo a debugging and tuning process (often referred to as "Software Quality Assurance" or SQA) prior to release, almost invariably bugs occur when a game or application is released to a wide audience. do. Unfortunately, it is extremely difficult for a software developer to find and identify many bugs after a product release. It is also very difficult for software developers to find bugs. Even if they become aware of bugs, much of the information that can be used to recognize the cause of the bug may be limited. For example, a user will call the game developer's customer service and leave a message stating that the PC will freeze as the screen begins to blink and turn solid blue when playing the game. This provides the SQA team with very little useful information for finding bugs. Some games or applications that go online can sometimes provide more information in some cases. For example, a "watchdog" process can sometimes be used to monitor a game or application for "crashes." When it crashes and uploads information to the SQA team via the Internet, the surveillance processor can collect statistics on the state of the game or application processor (eg, the state of the stack, memory usage, how far the game or application has progressed, etc.). have. However, in a complex game or application, it can take a very long time to read such information to determine exactly what the user will use at the time of the crash. Even so, it would be possible to determine what sequence of events led to the collision.

Another problem with PCs and game consoles concerns service issues that are very inconvenient for consumers. Also, service issues are typically associated with broken PCs or consoles. It is required to send it in a specific box for safe shipment, after which the PC or console will be incurred for repair if it is under warranty. Game or application software publishers may also lose sales (or May be affected by the use of online services).

1 illustrates a prior art video gaming system such as Sony PlayStation® 3, Microsoft Xbox 360®, Nintendo Wii ™, Windows-based personal computer or Apple Macintosh. Each of these systems has a central processing unit (CPU) to execute program code, typically a graphics processing unit (GPU) for performing advanced graphics operations, multiple forms of input for communicating with external devices and users. Contains / output (I / O). For simplicity, this configuration is shown combined with a single unit 100. In addition, the prior art video game system of FIG. 1 includes an optical media driver 104 (eg, a DVD-ROM drive); A hard drive 103 for storing video game program code and data; Network connection 105 for playing multiplayer games and downloading games, patches, demos, or other media; A random access memory (RAM) 101 for storing program code currently executed by the CPU / GPU 100; A game controller 106 for receiving input instructions from a user while playing a game; And a display device 102 (eg, SDTV / HDTV or computer monitor).

The prior art system shown in FIG. 1 suffers from some limitations. First, the optical drive 104 and hard drive 103 show much slower access speeds when compared to the RAM 101. When operating directly through the RAM 101, the CPU / GPU 100 is actually hard Processing more polygons per second than is possible when obtaining data and program code directly from drive 103 or optical drive 104. The reason is that the RAM 101 generally has a higher bandwidth and generally does not suffer from the long seek delay of the disk mechanism. However, only a limited amount of RAM is provided for this prior art system (e.g. 256-512 Mbytes). Thus, a "loading ..." sequence is often required in the continuous RAM 101, which is periodically filled with data for the next scene of the video game.

Some systems attempt to overlap loading of program code at the same time as gameplay, but this can only be done when there is a sequence of known events (e.g., when the car descends along the road, the structure for accessing buildings on the side of the road is Can be loaded while the car is running). For complex and / or rapid screen changes, this form of overlapping usually does not work. For example, if the user is at war and the RAM 101 is completely filled with data representing objects in the scene at that moment, if the user moves the scenes to the left to see an object that is not currently loaded in the RAM 101. Since there is no time to load new objects from the hard drive 103 or the optical media 104 into the RAM 101, there may be discontinuities in operation.

Another problem with the system of FIG. 1 arises due to the limitation of the storage capacity of hard drive 103 and optical media 104. Although disk storage devices can be manufactured with relatively large storage space (eg, 50 gigabytes or more), they still do not provide enough storage space for certain scenarios currently faced within video games. For example, as noted above, a football video game may allow a user to select from twelve teams, players, and stadiums throughout the world. Numerous document maps and environmental maps for each team, each player, and each stadium are needed to characterize the 3D appearance within the world (eg, each team has its own athletic shirt, each with its own document). Request a map).

One technique used to address this latter problem is for the game to pre-compute documents and environment maps as soon as they are selected by the user. This may include several computer-intensive processes, including stretching images, 3D mapping, shading, organizing data structures, and the like. As a result, there may be a delay for the user while the video game performs these operations. In principle, one way to reduce this delay is to perform all of these operations, including all changes to the team, player roster, and stadium when the game was first developed. The released version of the game then transfers the already processed data selected for the stadium selection, player roster, and team to the one or more servers or optical media 104 on the Internet that are downloaded by the user to the hard drive 103 via the Internet when selected. It will be included in all of these stored preprocessed data. As a practical matter, however, such preloaded data of all possible substitutions within gameplay can easily be several terabytes, which far exceeds the capacity of today's optical media devices. Moreover, the data for a given team, player roster, and stadium selection can easily be more than a few hundred megabytes of data. With a so-called 10 Mbps home network connection, it may take longer to download this data over the network connection 105 than to locally calculate the data.

Thus, the conventional game structure shown in FIG. 1 causes the user significant delays between major scene changes of complex games.

As shown in FIG. 1, another problem with the prior art is that over the years, video games tend to be more improved and require more CPU / GPU processing power. Thus, assuming that the amount of RAM is unlimited, video game hardware requirements will exceed the highest processing power available within such a system. As a result, users are required to upgrade their gaming hardware every few years to maintain pace (or to play new games at low quality). An important conclusion among the trends for better video games is that video game playing machines for home use are typically economically inefficient because their cost is usually determined by the requirements of the best performing games they can support. . For example, Xbox 360 can be used to play games such as "Gears of War," which require hundreds of megabytes of RAM, high performance CPUs, and GPUs, or Xbox 360 has a very low performance CPU and only a few kilobytes. It can also be used to play the Pac Man of the 1970s game, which requires RAM. In fact, Xbox 360 has enough computing power to host concurrent Pac-Man games at once.

Video game machines are typically off for most of the time of the week. According to a June 2006 Nielsen Entertainment study of active gamers 13 years of age and older, on average, active gamers spend 14 hours per week playing console video games, or 12% of the total time of the week. This means that the average video game console will not run for 88% of the time, which means inefficient use of expensive resources. This is particularly important as video game consoles are often subsidized by suppliers to lower consumer prices (with the expectation that subsidies can be earned again by royalties from future video game software purchases).

Video game consoles also incur costs associated with almost any consumer electronic device. For example, the electronics and mechanism of the systems need to be enclosed and housing. The manufacturer needs to provide a service certificate. Retailers who sell the system need to collect profits from any sales of the system and / or sales of video game software. All of these factors add to the cost of a video game console, which may be subsidized by the manufacturer, passed on to the consumer, or both.

Moreover, piracy is a major problem in the video game industry. The security mechanisms used on virtually all major video gaming systems are "cracked" over the years, resulting in unauthorized piracy of video games. For example, the Xbox 360 security system was cracked in July 2006, and users can now download pirated copies online. Downloadable (eg games for PC or Mac) games are partially weak in piracy. In some parts of the world where piracy is weakly policyd, there is essentially no viable market for standalone video game software because users can buy pirated copies as easily as legitimate at a fraction of the cost. Also, as the cost of game consoles accounts for a high percentage of income in many parts of the world, only a few people can afford to buy gaming systems in the technology, even if infringement is controlled.

Moreover, the game market used reduces revenue for the video game industry. When a user becomes bored with a game, they can sell the game to a store that can sell the game back to other users. This unauthorized but common practice significantly reduces the revenue for game publishers. Similarly, a 50% drop in on-demand sales usually occurs when there are platform changes every few years. This is because users stop buying games for older platforms when a newer version platform is about to be released (eg, when Playstation 3 is about to be released, people stop buying Playstation 2 games). Overall, the decrease in sales and the increasing development costs associated with the new platform can have a very significant adverse impact on the suitability of game developers.

New game consoles are also very expensive. The Xbox 360, Nintendo Wii, and Sony Playstation 3 all sell for hundreds of dollars. High-power PC gaming systems can cost more than $ 8000. This represents a significant investment for users, especially given the fact that many systems were sold to children and that hardware would not be available in a few years.

One approach to anticipated problems is online gaming in which gaming program code and data are hosted on servers and transmitted as on-demand client machines as compressed video and audio streamed across digital broadband networks. Certain companies, such as Finland's G-Cluster (now a subsidiary of Japan's SOFTBANK Broadmedia), now offer these online services. Similar gaming services are provided by DSL and cable television providers and are available in the local network, such as the network in hotels. The main weakness of these systems is the issue of latency, i.e., the time it takes for the signal to travel to and from the gamer server, which is typically located within the operator's "head-end." Fast action video games (also known as "twitch" video games) require a very low latency between the time a user performs an action with the game controller and the time the display screen updates the result of the user's action. do. Low-latency is necessary for the user to have the perception of the game responding "instantly". Users may be satisfied with different latency intervals depending on user skill level and game type. For example, a latency of 100 ms is quite good for slow casual games (such as backgammon) or slow-action role playing games, but for fast action games, a latency greater than 70 or 80 ms can be It can cause performance to be worse in the game, so it is unacceptable. For example, in games where fast response time is required, there is a steep slope in accuracy as the latency increases from 50 to 100 ms.

When a game or application server is installed in close proximity, the network path to the controlled network environment, or user, can predict and / or tolerate bandwidth peaks, which can deliver latency both in terms of maximum latency and consistency of latency. It is much easier to control (e.g., so the user observes steady motion from digital video streaming over the network). This level of control can be achieved between the cable TV network headend and the home of the cable TV subscriber, from the DSL central office to the DSL subscriber home, or from a server or user in a commercial office LAN environment. It is also possible to obtain specially-grade point-to-point personal connections between businesses with guaranteed bandwidth and latency. However, in a game or application system that hosts games in a server center connected to the general Internet and then streams compressed video to the user via a broadband connection, latency is caused by many factors and is critical in the deployment of conventional systems. Cause restrictions.

In a typical broadband connection home, a user may have a DSL or cable modem for broadband service. These broadband services typically generate 25ms round-trip latency (and more than double) between the user's home and the Internet. Moreover, there are round trip latencies that occur from routing data to server centers over the Internet. Latency over the Internet changes based on delays that occur as data is given and routed. In addition to routing delays, round-trip latency also occurs due to the speed of light transmitted through optical fibers that interconnect most of the Internet. For example, about 22 ms each 1000 miles is generated with reciprocal latency due to the speed of light through the optical fiber and other overhead.

Additional latency may occur due to the data rate of data streamed over the Internet. For example, if a user has a DSL service that is sold as a "6 Mbps DSL service," in fact, the user will probably have something less than 5 Mbps upstream, and probably a lot of things like congestion during peak load times in the Digital Subscriber Line Access Multiplexer (DSLAM). You will see the connection drop off periodically because of the factors. A similar issue is that if there is congestion at a local shared coaxial cable looped through another location or adjacent one in the cable modem system network, then a much smaller "6Mbps cable modem service". It is possible to reduce the transmission rate of cable modems using connections sold as If data packets are streamed in one direction in the User Datagram Protocol (UDP) format from the server center via this connection at a stable rate of 4 Mbps, when all works well, the data packets will not introduce additional latency. If it will pass, but there is congestion (or other obstacles), only 3.5 Mbps can be used to stream data to the user, so in a typical situation, packets will drop, causing lost data, or packets they It will queue at the congestion point until it can be transmitted with incurring additional latency. Different points of congestion have different queuing capacity to catch delayed packets, so in some cases, packets that cannot be made through congestion will drop immediately. In other cases, several megabits of data will be queued up and eventually sent. However, in almost all cases, the queues at congestion points have capacity limits, and if these limits are exceeded, the queues may overflow and packets will drop. Thus, in order to avoid the occurrence of additional latency (or loss of worse packets), it is necessary to avoid exceeding the rate capacity from the game or application server to the user.

Latency is also generated by the time required to compress the video within the server and to stretch the video within the client device. Latency is further generated while calculating the next frame for the video game in progress on the server to be displayed. Currently available video compression algorithms suffer from either high data rates or high latency. For example, motion JPEG is an intraframe-only lossy compression algorithm characterized by low latency. Each frame of video is compressed independently of the other frames of the video. When a client device receives a frame of compressed motion JPEG video, it will immediately stretch the frame and display it, resulting in very low latency. However, because each frame is compressed separately, it is impossible for the algorithm to take advantage of the similarity between successive frames, and consequently the intraframe-only video compression algorithm suffers from a very high bit rate. For example, 640x480 motion JPEG video at 60fps (frames per second) may require 40Mbps or more data. This high bit rate for such low resolution video windows can be very expensive in many broadcast applications (and certainly for most consumer Internet based applications). Moreover, because each frame is compressed independently, artifacts within a frame that may result from lossy compression appear to appear in different places within successive frames. This results in the viewer appearing as moving visual artifacts when the video is stretched.

As they are used in prior art configurations, other compression algorithms such as Microsoft Corporation's MPEG2, H.264 or VC9 can achieve high compression ratios but at a high latency cost. These algorithms perform intraframe-only compression of the frame. Such a frame is known as a key frame (typically called an "I" frame). Thus, these algorithms typically compare an I frame with previous frames and successive frames. Rather than compressing previous frames and successive frames independently, the algorithm is better to determine changing in the image from I frame to previous and successive frames, thus "B" for changes that precede the I frame. It stores these changes, called a "P" frame, for the frame, and the changes that follow the I frame. This results in a much lower bit rate than intraframe-only compression. However, it typically appears at a higher latency cost. I frames are typically much larger (often 10 times larger) than B or P frames, and consequently take longer to propagate at a given rate. For example, if I frames are 10 times the size of B and P frames, there are 29 B frames + 30 P frames = 59 interframes for every single I frame, or each " Group of Frames; Total 60 frames per GOP). Thus, at 60 fps, there is one 60-frame GOP per second.

Imagine the transport channel has a maximum rate of 2Mbps. To achieve the highest quality data stream in the channel, the compression algorithm will produce a 2 Mbps data stream, given the above ratio, which is equal to 2 megabits (Mb) / (59 + 10) = 30,394 bits / intraframe. Will result in 303,935 bit / I frames. When a compressed video stream is received by the decompression algorithm, each frame needs to be displayed and decompressed at regular intervals (eg, 60 fps) in order for the video to run stably. To achieve this result, if a frame is subject to transmit latency, all frames need to be delayed by at least that latency, and thus the worst-case frame latency will define latency during all video frames. . Because I frames are the largest, they generate the longest transmission latency, and the entire I frame will be received before the I frame is stretched and displayed (or an interframe dependent on the I frame). Given a channel rate of 2Mbps, it would take 303,935 / 2Mb = 145ms to transmit an I frame.

As described above using a large proportion of the bandwidth of the transport channel, an interframe video compression system will be subject to long latency due to the large size of the I frame proportional to the average size of the frame. Or, to solve it the other way, while prior art interframe compression algorithms achieve a lower average per frame rate than intraframe-only compression algorithms (e.g., 2 Mbps to 40 Mbps), still they are still due to large I frames Suffer a high peak per frame rate (e.g., 303,935 * 60 = 18.2 Mbps). Keep in mind, through the above analysis, assume that both P and B frames are much smaller than I frames. While this is generally true, it is not true for frames with high image complexity that are not associated with scene changes, a lot of motion, or previous frames. In such a situation, P or B frames can be as large as I frames (if the P or B frames are larger than I frames, sophisticated compression algorithms will typically "force" I frames and convert P or B frames to I frames). Will change to). Thus, peaks in I frame-size rates can occur at any instant in the digital video stream. Thus, with compressed video, when the average video rate approaches the rate capacity of the transmission channels (if frequent, given a high rate requirement for the video), high peak rates from I frames or large P or B frames are high. Cause frame latency.

Naturally, the discussion above only characterizes the compression algorithm latency generated by large B, P or I frames in the GOP. If B frames are used, the latency is much higher. The reason is that before the B frame can be displayed, all of the B frames and the B frames after the I frame are received. Thus, within a group of picture (GOP) of a sequence such as BBBBBIPPPPPBBBBBIPPPPP, where there are five B frames before each I frame, the first B frames are video until subsequent B frames and I frames are received. It cannot be displayed by the decompressor. So if the video is streamed at 60fps (i.e. 16.67ms / frame), then 5 B frames and I frame take 16.67 * 6 = 100ms to receive, no matter how fast the channel bandwidth is, it has only 5 B frames. . Compressed video sequences with 30 B frames are very common. And at low channel bandwidths, such as 2 Mbps, the latency effect caused by the size of the I frame adds significantly to the latency effect because it waits for B frames to arrive. Thus, on a 2 Mbps channel, if you have a huge number of B frames, it is very easy to exceed the 500 ms latency using conventional video compression techniques. If B frames are not used (at the cost of a low compression ratio for a given quality level), as mentioned above, B frame latency does not occur, and only the latency caused by peak frame sizes still occurs.

The problem is exacerbated by the important nature of many video games. Video compression algorithms using the GOP structure described above have been highly optimized for the use of live motion or picture material intended for passive viewing. Typically, if the camera or scene is moving too suddenly, if the video or movie material is (a) typically uncomfortable to watch and (b) it is visible, the average viewer may not be able to closely follow the action when the camera moves suddenly. Cameras (whether they are real cameras or virtual cameras for computer-made animations) and the scenes are relatively stable (eg, when a child blows out candles on a birthday cake and suddenly returns to the cake and blows again If the camera falls when the camera is turned off, viewers typically focus on the child and the cake, and when the camera suddenly moves, the short pause is ignored). In the case of a video interview, or video teleconferencing, the camera can be in a fixed position and there is no movement at all, resulting in very small data peaks. However, 3D high action video games are characterized by constant motion (eg, when the whole frame is in fast motion while playing, thinking about 3D racing, or when the virtual camera always moves suddenly, who shoots the first person). Think of it). Such video games result in a frame sequence with large and frequent peaks that the user needs to clearly show what happens during these sudden actions. As in this case, compression artifacts are very bad in 3D high action video games. Thus, the video output of many video games, by their nature, produces a compressed video stream with very high and frequent peaks.

If a user of a fast-action video game has high latency and a small endurance for the cause of all of the given latency, then there is a limitation in server-hosted video games that stream video to the Internet. Moreover, users of applications that require a high degree of interaction experience similar limitations if the applications are hosted on the general Internet and stream video. In order for the route and distance from the client device to the server to be controlled to minimize latency, and to allow peaks to be accommodated without latency, these services are provided by the hosting server in the head-end (in the case of cable broadband) or in the central office. (In the case of DSL), or directly within a LAN (or on a specially-graded personal connection) in commercial settings. LANs with adequate bandwidth (typically with speeds of 100 Mbps-1 Gbps) and leased lines can typically support peak bandwidth requirements (eg, 18 Mbps peak bandwidth is a small portion of 100 Mbps LAN capacity).

Peak bandwidth requirements can also be accommodated by residential broadband infrastructure if special capacity is made. For example, on a cable TV system, digital video traffic may have a dedicated bandwidth that can handle peaks as large as I frames. And, on a DSL system, a higher speed DSL modem can be supplied allowing high peaks, or a special-grade connection can be supplied to handle higher data rates. However, traditional cable modem and DSL infrastructures attached to the general Internet have much less error tolerance for peak bandwidth requirements for compressed video. Thus, an online service that hosts a video game or application in server centers remote from a client device and streams compressed video output over the Internet over traditional home broadband connections is particularly demanding for games that require very low latency. And applications (eg, first-person shooters and other multi-user, interactive action games, or applications requiring fast response times) suffer from significant latency and peak bandwidth limitations.

SUMMARY OF THE INVENTION The present invention has been made in view of the foregoing, and an object thereof is to provide a system and method for storing program code and data in an application hosting center.

This object according to the invention comprises the steps of: dividing program code and / or data used to execute an online application into a first type and a second type; Storing the first type of program code and data in a first type of memory that provides relatively low latency memory access; Storing the second type of program code and data in a second type of memory that provides relatively higher latency memory access than the first type of memory; Retrieving program code and data from the first and second types of memory in response to a client request to execute an online application; And transmitting to the client a streaming interactive video stream representing an image generated by the application.

A system and method for storing program code and data in such an application hosting center can be provided.

1 illustrates a prior art video gaming program architecture;
2A and 2B illustrate a high level system architecture according to one embodiment;
3 is a diagram showing actual, speed, and required transmission rate for communication between a client and a server;
4A illustrates a hosting service and an adopted client, according to one embodiment;
4B illustrates an exemplary latency associated with communication between a client and a hosting service;
4C is a diagram of a client device according to one embodiment;
4d illustrates a client device according to another embodiment;
4E illustrates an example of a block diagram of a client service in FIG. 4C;
4F illustrates an example of a block diagram of a client service in FIG. 4D;
5 illustrates an example of a form of video compression that may be employed in accordance with one embodiment;
6A illustrates an example of a form of video compression that may be employed in another embodiment;
6b shows a peak in bit rate associated with low complexity transmission, low motion video sequence,
6C shows a peak in transmission rate associated with high complexity transmission, high motion video sequence,
7A, 7B illustrate an example of a video compression technique employed in one embodiment;
8 illustrates a further example of a video compression technique employed in one embodiment;
9A-9C show examples of techniques employed in one embodiment for relaxed rate peaks,
10A and 10B illustrate one embodiment for efficiently packing an image tile in a packet.
11A-11D illustrate an embodiment employing forward error correction techniques,
12 illustrates one embodiment using a multi-core processing unit for compression;
13A and 13B illustrate geographical positioning and communication with a hosting service according to various embodiments;
14 illustrates exemplary latency associated with communication between a client and a hosting service;
15 shows an example of a hosting service server center architecture,
16 shows an example of a screenshot of one embodiment of a user interface including multiple live video windows;
17 illustrates the user interface of FIG. 16 in accordance with the selection of a particular video window;
18 illustrates the user interface of FIG. 17 in accordance with an enlargement of a particular video window of full screen size;
19 shows an example of common user video data overlaid on the screen of a multiplayer game;
20 shows an example of a user page for a game player in a hosting service;
21 is a diagram showing an example of a 3D interactive advertisement,
FIG. 22 shows an example of a sequence of steps of producing a photoreal image with a textured surface from surface capture of live performance; FIG.
23 shows an example of a user interface page allowing the selection of linear media content;
FIG. 24 shows a graph of the speed of connection before the web page is live.

The invention will be more fully understood from the accompanying drawings and the following detailed description. However, the particular embodiments shown are not intended to limit the technical spirit of the disclosed invention, but are for understanding and explanation only.

In the following detailed description, specific details are set such as device type, system configuration, and communication method in order to provide an understanding of the present disclosure. However, one of ordinary skill in the relevant art will recognize that these specific details may not be required to practice the described embodiments.

2A and 2B show that video games and software applications are hosted by hosting service 210 and user premises 211 (“user premises”) over the Internet 206 (other public or private networks) under a subscription service. Means the location of the user, including the outside, if a mobile device is used). The client device 205 is a general purpose computer, such as Microsoft Windows, which is wired or wirelessly connected to the Internet, or a Linux-based PC, or a Macintosh computer of Apple Inc., or is connected to a video or an internal or external display device 222 or video. And a client device such as a set-top box (connected with the Internet wired or wirelessly) that outputs audio to a monitor or TV set 222, possibly a mobile device connected to the Internet wired or wirelessly.

Some of these devices may have their own user input devices (eg, keyboards, buttons, touchscreens, trackpads, or inertial sensing bars, video capture cameras, and / or motion tracking cameras), or they may be wired or wirelessly. A connected internal input device 221 (eg, keyboard, mouse, game controller, inertial sense bar, video capture camera and / or motion tracking camera, etc.) is used. As described below in more detail, hosting service 210 includes servers of varying levels of performance, including high power CPU / GPU processing performance. During the use of an application or the playing of a game at the hosting server 210, the home or office client device 205 receives keyboard and / or controller input from the user, which then receives the game or application software (through the Internet 206). For example, if the user presses a button that directs a character on the screen to move to the right, then the game program creates a sequence of video images that show the character moving to the right). The controller input is sent to the hosting service 210 which generates the frame and executes the gaming program code in response. This sequence of video images is compressed using a low-latency video compressor, and the hosting service 210 sends the low-latency video stream over the internet 206. The home or office client device decodes the compressed video stream and renders the decompressed video image on the monitor or TV. As a result, the computing and graphics hardware requirements of client device 205 are significantly omitted. The client 205 only needs to have the power to send keyboard / controller input to the internet 206, decompress and decompress the compressed video stream received from the internet 206, which is In fact, a personal computer is capable of running software on its CPU (for example, Intel's Core Duo CPUs running at approximately 2 GHz use compressors such as Windows Media VC9 and H.264 to compress encoded 720p HDTVs). Can be solved). And, in the case of some client devices, the dedicated chip may also perform video compression for such standards in real time at a lower cost and less power consumption than a general purpose PC as required for a modern PC. In order to perform the function of remarkably sending controller input and decompressing video, the home client device 205 may comprise an optical drive, hard drive, or any special graphics processing unit (such as the prior art video game system shown in FIG. 1). GPU).

As gaming and application software become more complex and more photorealistic, they will require higher performance CPUs, GPUs, more RAM, larger and faster disk drives, and computing power in the hosting service 210 will be continuously upgraded, The end user will not be required to update the home or office client platform 205 because its processing needs will remain fixed for display resolution and frame rate with an algorithm to decompress existing video. Thus, hardware limitations and compatibility issues no longer exist in the systems shown in FIGS. 2A and 2B today.

Moreover, since the game and application software run only on the server of the hosting service 210, the user's home or office (where "office") would include any non-resident setting of any qualification, including, for example, a school room. Never copy a game or application software (either in the form of optical media, or downloaded software). This significantly mitigates the likelihood that a game or application is pirated, as well as the probability of a valuable database that can be used by the pirated game or application. Indeed, if specialized servers are required to play ineffective game or application software for use in the home office, it may work in the home or office even though piracy of the game or application software has been obtained. You will not be able to.

In one embodiment, hosting service 210 is a game or application software developer (typically a software development company, game or movie studio or game or application) that develops a video game and designs a game that can be executed on hosting service 210. It is referred to as a software publisher; it provides software development tools for 220. Such tools allow developers to take advantage of features of hosting services that are not normally available on standard PCs or game consoles (eg, complex geometry ("geometry") otherwise polygons defined here as 3D datasets). Very fast access to a very large database of textures, rigging, lighting, behavior and other elements and parameters).

Other business models are possible under this architecture. Under one model, hosting service 210 collects a subscription fee from an end user and pays a loyalty to developer 220, as shown in FIG. 2A. In the alternative implementation shown in FIG. 2B, the developer 220 collects the subscription fee directly from the user and pays the hosting service 210 for hosting the game or application content. This fundamental theory is not limited to any specific business model for providing online gaming or application hosting.

Compressed video features ( Compressed Video Characterisitics )

As discussed above, one prominent problem with service online provision of video game services or application software is latency. Latency of 70-80 ms (from the point of input device actuated by the user to the point of response displayed on the display device) is an upper limit for games and applications requiring fast response time. However, this is extremely difficult to achieve with respect to the architecture shown in FIGS. 2A and 2B due to many practical physical constraints.

As shown in Figure 3, when a user subscribes to an Internet service, the connection is typically rated at a nominal maximum transfer rate 301 to the user's home or office. Depending on the provider's policy and the capabilities of the transmitting equipment, the maximum rate is strictly faster or slower, but typically the actual rate available is slower for one of many other reasons. For example, in a DSL central office or local cable modem loop there may be too much network traffic, there may be noise in the cable causing dropped packets, or the provider will set a maximum bit per user per month. Currently, the maximum downstream rates for cable and DSL services typically range from several hundred kilobits per second (Kbps) to 30 Mbps. Cellular services are typically limited to hundreds of Kbps of downstream data. However, the speed of broadband services and the number of users who subscribe to broadband services will increase dramatically over time. Currently, some analysts estimate that 33% of US broadband subscribers have downstream speeds above 2Mbps. For example, some analysts predict that by 2010, more than 85% of US broadband subscribers will have data rates above 2Mbps.

As shown in FIG. 3, the actual maximum available rate 302 will vary over time. Thus, at low latency, the online game or application software associated with it is difficult to predict the actual rate available at a particular video stream. If the rate 303 is greater than the actual maximum available rate 302, then a number of frames-per-second fps at a given resolution (e.g. 640 x 480 at 60 fps) for any number of complex scenes Required to maintain a given level of quality will result in distorted / lost images and lost data on the user's video screen. Other services temporarily buffer additional packets (e.g., queue up), and-increase the latency, resulting in unacceptable results for many video games and applications, by providing packets to the client at the available rates. Brings about. Finally, some Internet service providers, such as the denial of service attacks (a well-known technology user used by hackers to prevent network connections), increase the transmission rate due to malicious attacks, and disconnect users from the Internet for a certain period of time. Will see. Thus, the embodiment described herein goes through steps to ensure that the rate required for a video game does not exceed the maximum available rate.

Hosting Service Architecture ( Hosting Service Architecture )

4A illustrates an architecture of hosting service 210 according to one embodiment. Hosting service 210 may be located in a single server center or distributed across multiple server centers (one or more servers to provide low-latency connections to users with low-latency paths to any server center than other users). To provide redundancy in the event of a center failure and to provide a balanced load among multiple users). The hosting service 210 will eventually include hundreds of thousands or millions of servers 402 providing a very large user base. The hosting service control system 401 controls the hosting service 201 as a whole and controls the router, server, video compression system, billing and accounting system, and the like. In one embodiment, the hosting service control system 401 runs on a distributed processing Linux based system tied to a RAID array used to store a database for system statistics, server information, and user information. In the foregoing detailed description, various operations performed by the hosting service 210 are initiated and controlled by the hosting service control system 401, unless viewed as a result of other specific systems.

Hosting service 210 includes a number of servers 402, such as those currently available from Intel, IBM, Hewlett Packard, and others. Alternatively, server 402 may be assembled into a customary configuration of components, or eventually integrated so that the entire server runs as a single chip. Although this diagram shows a small number of servers 402 for purposes of illustration, there will be at least one server 402 or more than millions of servers in a practical deployment. The server 402 has the same method (an example of any of the configuration parameters, with the same CPU type and performance; with or without a GPU, if there is a GPU, the same GPU type and performance; with the same number of CPUs and GPUs; the same Have a quantity of form / speed of RAM; and all set to the same RAM configuration, or various subsets of the server 402 may be configured in the same configuration (e.g., 25% of the servers may be configured in a particular way). 50% may be configured in a different manner, and 25% may already be configured in another manner), or all servers 402 may be different.

In one embodiment, server 402 is diskless, e.g., itself a local mass storage (a mass storage that is a means for performing optical or magnetic storage or semiconductor based storage such as flash memory or other similar functions). Rather than having, each server has access to shared mass storage through a fast backplane or network connection. In one embodiment, this fast connection is a Storage Area Network (403) connected to a series of Redundant Arrays of Independent Disks (RAID) 405 having connections between devices running using Gigabit Ethernet. As is known in the art, the SAN 403 is coupled with many RAID arrays 405, resulting in extremely high bandwidth-approaching or potentially exceeding the bandwidth available in the RAM currently used in game consoles and PCs. Will be used. And RAID arrays based on rotating media, such as magnetic media, often have significant seek time access latency, while RAID arrays based on semiconductor storage can be implemented with lower access latency. In other configurations, some or all of the servers 402 provide some or all of their own mass storage locally. For example, server 402 will store frequently accessed information such as copies of video games or applications and its operating system in its operating system and low-latency local flash-based storage, but based on the frequency The SAN will be used to access the RAID array 405 based on game state information.

Moreover, in one embodiment, hosting service 210 employs low-latency video compression logic 404, described in detail below. Video compression logic 404 may be executed in software, hardware, or some combination thereof (any embodiment described below). Video compression logic 404 includes logic for audio compression as well as visual material.

In operation, while playing a video game or using an application in the user premises 211 via a keyboard, mouse, game controller, or other input device 421, the control signal logic 413 at the client 415 is stored by the user. Control signals 406a, 406b (typically in the form of UDP packets) indicating button presses (and other forms of user input) actuated by < RTI ID = 0.0 > Control signals from the existing user are sent to the appropriate server (or servers, if multiple servers respond to the user's input device). As shown in FIG. 4A, the control signal 406a will be sent to the server 402 via the SAN. Alternatively or additionally, control signal 406b may be sent directly to server 402 via a hosting service network (eg, an Ethernet based local area network). Regardless of how they are transmitted, the server or servers execute game or application software that responds to control signals 406a and 406b. Although not shown in FIG. 4A, various networking elements, such as firewalls and / or gateways, may include user premises 211 between the edge of hosting service 210 and / or between Internet 410 and home or office client 415. It will handle the incoming and outgoing traffic at the edge of). The graphics and audio output of the game or application software being executed, such as a new continuous video image, is provided to the low-latency video compression logic 404 that compresses the sequence of video images according to low-latency video compression techniques as described herein. And a compressed video stream, typically compressed or uncompressed audio, via the Internet 410 (or via an optimized high speed network service through the general Internet, as described below). Let's go back to). Low-latency video decompression logic 412 at client 415 then decompresses the video and audio streams and renders the decompressed video stream, and typically plays the decompressed audio system at display device 422. . Alternatively, the audio may be played in a speaker that is not separated or separated from the display device 422, noting that the input device 421 and the display device 422 are shown independently in FIGS. 2A and 2B. They will be integrated into a client device, such as a portable computer or mobile device.

The home or office client 415 (previously described as the home or office client 205 in FIGS. 2A and 2B) will be a very inexpensive and low power device with very limited computing or graphics performance and is much more limited or local. Will not have large storage. In contrast, each server 402 combined with SAN 403 and multiple RAID 405 may be an exceptionally high performance computing system, and in fact if multiple servers are used in concert in a parallel processing configuration, There are very few limits to the many computing and graphics processing power that can be purchased to withstand. And because of low-latency video compression 404 and low-latency video compression 412, perceptually to the user, the computing power of server 402 will be provided to the user. When the user presses a button on the input device 421, the image on the display 422 is updated in response to the button press, with no significant delay, as if the game or application software were operated locally. Thus, a very low performance computer or home or office client 415 with only cheap chips executes low-latency video stretch and control signal logic 413, and the user is efficient random computing power at a remote location as locally available. Is provided. This presents the user with the most advanced, processor-intensive (typically new) video games and highest performance applications.

4C shows a very basic and inexpensive home or office client device 465. This device is an embodiment of a home or office client 415 from FIGS. 4A and 4B. It is approximately two inches long. It has an Ethernet jack 462 that interfaces with an Ethernet cable with Power over Ethernet (PoE), from which a connection between power and the Internet is obtained. It can operate NAT within a network that supports NAT (Network Address Translation). In an office environment, many new Ethernet switches have PoE and take PoE directly to an Ethernet jack in the office. In this situation, all that is required is an Ethernet cable from the wall jack to the client 465. If the available Ethernet connection does not carry power (eg, with a DSL or cable modem at home, but not PoE), an inexpensive available wall "brick that allows output Ethernet with PoE and powerless Ethernet cables ( bricks) "(eg power supplies).

Client 465 includes control signal logic 413 (FIG. 4A) coupled with a Bluetooth wireless interface, which interfaces with a Bluetooth input device 479 such as a keyboard, mouse, game controller and / or microphone and / or headset. do. In addition, one embodiment of client 465 is a display device 468 capable of supporting 120 fps video and audio with a pair of shuttered glasses (typically via infrared) to alternatively shutter one eye. ) Can output video at 120 fps. The effect perceived by the user is a "jumps out" stereoscopic 3D image of the display screen. The display device 468 supporting such an operation is the Samsung HL-T5076S. By separating the video streams for each eye, in one embodiment two independent video streams are compressed by the hosting service 210, the frames are fitted in time, and the frames are two independent extensions in the client 465. Elongated by treatment.

The client 465 also includes low-latency video decompression logic 412, which decompresses incoming video and audio and provides video and audio to a TV via a high-definition multimedia interface (HDMI). Output via a connector 463 plugged into a Standard Definition Television) or HDTV (High Definition Television), or a monitor 468 that supports HDMI. If your monitor 468 does not support HDMI, HDMI to DVI (Digital Visual Interface) can be used, but audio will be lost. Under the HDMI standard, display performance 464 (e.g., supported resolution, frame rate) communicates with display device 468, and then this information is communicated to hosting service 210 via Internet connection 462 and the display device. You can stream compressed video in an appropriate format.

FIG. 4D shows the same, except that the home or office client device 475 further has an external interface with the home or office client device 465 shown in FIG. 4C. In addition, the client 475 can be obtained from an external power supply adapter (not shown) that plugs into the wall as well as PoE to obtain power. If client 457 uses a USB input, video camera 477 provides compressed video to client 475, which is uploaded by client 475 for hosting service 210 in the use described below. . Consisting of a camera 477 is a low-latency compressor using the compression technique described below.

By adding an Ethernet connector for the Internet connection, the client 475 also has an 802.11g air interface for the Internet. Both interfaces can use NAT in networks that support NAT.

In addition to having HDMI connectors for video and audio outputs, the client 475 also has dual link DVI-I connectors that include analog outputs (and standard adapter cables will provide VGA outputs). It also has analog outputs for composite video and S-video.

For audio, the client 475 has a left / right analog stereo RCA jack and a TOSLINK output for digital audio output.

In addition to the Bluetooth wireless interface for the input device 479, a USB jack is also required to interface with the input device.

4E illustrates one embodiment of an internal architecture of client 465. All or some of the devices shown in the diagram may be implemented in a custom ASIC or some individual device, a Field Programmable Logic Array, which is a custom design or specification.

Ethernet 497 with PoE is attached to Ethernet interface 481. Power 499 is derived from Ethernet 497 with PoE and is connected to the rest of the device at client 465. Bus 480 is a common bus for communication between devices.

A control CPU 483 (some small CPU, such as the MIPS R4000 CPU series at 100 MHz with embedded RAM, is generally suitable) to run a small client control application from Flash 476 also runs a protocol stack for the network. Communicate with hosting service 210 and configure all devices at client 465. It also handles the interface with the input device 469 and, if necessary, sends the packet back to the hosting service 210 with user controller data protected by forward error correction. In addition, the control CPU 483 monitors packet traffic (eg, packets also timestamp their arrival if packets are lost or delayed). This information is sent back to the hosting service 210 to constantly monitor the network connection and adjust it accordingly. Flash memory 476 initially loads a control program for control CPU 483 and a serial number unique to a particular client 465 unit. This serial number is allowed to uniquely identify the client 465 unit.

The Bluetooth interface 484 communicates with the input device 469 wirelessly through an antenna inside the client 465.

Video stretcher 486 is a low-latency video stretcher configured to perform the video stretching described herein. Many video stretching devices exist with an Intellectual Property (IP) or off-the-shelf of the design that can be integrated into an FPGA or custom ASIC. One company that offers IP for H.264 decoders is Ocean Logic of Manly, New South Wales, Australia. The advantage of using IP is that the compression technique used here does not conform to compression standards. Some standard expanders are sufficient to be flexible here to accommodate compression techniques, while others do not. But with IP, it has all the flexibility to redesign the stretcher as required.

The output of the video extender is connected to the video output subsystem 487, which is connected to the video for video output of the HDMI interface 490.

The audio stretching subsystem 488 may use a standard audio extender available, or it may be implemented as an IP, or the audio extender may be executed within the control processor 483, which may run, for example, a Vorbis audio marketer.

The device for performing audio stretching is coupled with an audio output subsystem 489 which is coupled with audio for audio output of the HDMI interface 490.

4F illustrates one embodiment of an internal architecture of client 475. The architecture is the same as that of the client 465 except for the optional external DC power and added interface from the wall power plug adapter and, if so used, replaces the power coming from the Ethernet PoE 497. The purpose in common with the client 465 will not be repeated below, but the added purpose is described below.

The CPU 483 communicates with the additional device and constitutes the additional device.

The WiFi subsystem 482 provides wireless Internet that alternatively connects to the Ethernet 497 via an antenna. WiFi subsystem 485 may be used from a wide range of manufacturers, including Atheros Communication, Santa Clara, California.

USB subsystem 485 provides alternative to Bluetooth communication for wired USB input 479. The USB subsystem is quite standard and fairly available in FPGAs and ASICs, as well as often designed as a standardized device to perform other functions such as video stretching.

The video output subsystem 487 produces a wider range of video outputs within the client 465. In addition to the video output provided by HDMI 490, DVI-I 491, S-video 492, and composite video 493 are provided. In addition, when the DVI-I 491 interface is used for digital video, the display capability 464 is passed back from the display device to the control CPU 483 to inform the hosting service 210 of the display device 478 capability. Can be. All interfaces provided by video output subsystem 487 are fairly standard interfaces and are available in quite a few forms.

Audio output subsystem 489 outputs digital audio through digital interface 494 and outputs audio in analog form via stereo analog interface 495.

round trip Latency  analysis( Round - Trip Latency Analysis )

Of course, as can be seen from the advantages of the foregoing paragraphs, the reciprocal latency between the user's actions using the input device 421 and showing the result of the action on the display device 420 is no greater than 70-80 ms. This latency of the user premises 211 All elements in the path from the input device 421 to the hosting service 210 and back to the user premises 211 for the display device 422 should be considered. 4B illustrates various configurations and networks through which signals must travel, and these elements and networks are timelines listing exemplary latencies that can be expected in practical implementation. It should be noted that FIG. 4B is simplified so that only critical path routing is shown. The routing of other data used for other features of the system is described below. Both arrows (eg, arrow 453) point to reciprocation latency, one arrow (eg, arrow 457) point to one-way latency, and "~" represent approximate measurements. This would indicate that there would be real-world situations that the listed latency could not achieve, but in many cases in the United States, when used to connect a DSL and cable modem with the user premises 211, such latency is described in the following section. Can be achieved under the same circumstances. In addition, cellular wireless connections to the Internet will certainly operate in the illustrated system, whereas most current US cellular data systems (such as EVDO) generate very high latency and cannot achieve the latency shown in FIG. 4B. will be. However, this fundamental theory will be implemented in future cellular technologies with the ability to perform this level of latency.

Once started from the input device 421 in the user premises 211, the user operates the input device 421, and the user control signal may be a client 415 (which may be a basic device such as a set top box, or it may be a PC or May be software or hardware running on another device, such as a mobile device), packetized, and the packet is given a destination address to reach the hosting service 210. The packet also contains information indicating the user when a control signal comes in. Control signal packet (s) are then sent to the WAN interface 442 through the firewall / router / Network Address Translation (NAT) device 443. The WAN interface 442 is an interface device provided to the user premises 211 by the user's Internet Service Provider (ISP). WAN interface 442 may be a cable or DSL modem, a WIMAX transceiver, a fiber transceiver, a cellular data interface, an Internet protocol interface on a powerline, or many other interfaces for the Internet. Moreover, firewall / router / NAT device 443 (and potential WAN interface 442) will be integrated with client 415. Such an example would be a mobile phone, including software for performing the functions of a home or office client 415 as well as a means of wirelessly connecting and transmitting to the Internet via some standard (eg, 802.11g).

WAN interface 442 is then referred to herein as "interconnection" for a user's Internet Service Provider (ISP), a facility that provides an interface between the WAN transport associated with user premises 211 and a general Internet or private network. Transmits a control signal, referred to as a "point of presence". The nature of the interconnection location will vary depending on the nature of the Internet service provided. In the case of a DSL, it will typically be the telephone company central office where the DSLAM is located. In the case of a cable modem, it will typically be a cable Multi-System Office head end. In the case of a cellular system, it will typically be a control room associated with the cellular tower. However, whatever the nature of the interconnection location, control signal packet (s) will be sent to the general Internet 410. The control signal packet (s) will then be most easily sent to the WAN interface 410 for the hosting service 210 via the fiber transceiver interface. WAN 441 will then send control signal packets to routing logic 409 (which can be executed in many different ways, including Ethernet switches and routing servers), which evaluates the user's address and for the given user. Send control signal (s) to the correct server 402.

The server 402 then takes the control signal as input for game or application software running on the server 402 and uses the control signal to process the next frame of the game or application. Once the next frame is generated, video and audio will be output from the server 402 to the video compressor 404. Video and audio will be output from the server 402 to the compressor 404 through various means. First, the compressor 404 will be built into part of the server 402, so compression will be performed locally within the server 402. Alternatively, video and / or audio may be output in packetized form over a network connection, such as a private connection between server 402 and video compressor 404 or an Ethernet connection over a shared network such as SAN 403. . Or, the video will be output from the server 402 via a video output connector, such as a DVI or VGA connector, and then captured by the video compressor 404. The audio is also output from the server 402 as analog audio encoded and digitized by the audio compression logic in the digital audio (eg, via the TOSLINK or S / PDIF connector) or in the video compressor 404.

Video compressor 404 captures the video frame and the audio generated during the frame time from server 402, and then the video compressor will compress the video and audio using the techniques described below. Once the video and audio are compressed, it is packetized to an address for returning back to the user's client 415, which is sent to the WAN interface 441, and then the video and audio packets are sent over the general Internet 410. Send video and audio packets to the user's ISP interconnection location 441, then send video and audio packets to the WAN interface 442 at the user's premises, and send the video to the firewall / router / NAT device 443. And send audio packets, and then send video and audio packets to client 415.

Client 415 extends video and audio, then displays video on display device 422 (or client's integrated display device), and integrates amplifier / speaker on display device 422 or on individual amplifiers / speakers or clients. Send audio to

For the user to recognize the entire process described without a perceptual lag, the round trip delay needs to be less than 70 or 80 ms. Any latency delay in the described round trip path is under the control of hosting service 210 and / or is not a user or anyone else. Nevertheless, based on the testing and analysis of a large number of real-world scenarios, the following is a rough measure.

One-way transmission time for sending control signal 451 is typically less than 1 ms, and round trip transmission through user premises 452 is typically accomplished using a firewall / router / NAT switch over readily available consumer grade Ethernet at approximately 1 ms. Can be. User ISPs vary widely in their round trip delay 453, but with DSL and cable modem providers, we can typically see between 10 and 25 ms. Round-trip latency in the general Internet can vary greatly depending on how much traffic is sent and what failures have occurred in the transmission, but typically the general Internet provides a fair and optimal route (and this issue). Latency is generally determined by the speed of light through the fiber and the distance given to the destination. As further described below, the maximum distance we expect from the hosting premises 210 to fall from the user premises 211 is established to be roughly 1000 miles. At 1000 miles (2000 miles round trip), the actual time to send a signal over the Internet is approximately 22ms. WAN interface 441 for hosting service 210 is typically a high speed interface of commercial grade fiber with negligible latency. Thus, typical internet latency 454 is typically between 1 and 10 ms. One-way routing 455 latency through hosting service 210 may be achieved in less than 1 ms. Server 402 will typically compute new frames for a game or application in less than one frame time (16.7 ms at 60 fps) and 16 ms is the maximum one-way latency 456 reasonable to use. In the optimized hardware implementation of the video compression and audio compression algorithms described herein, compression 457 can be completed within 1 ms. In the less optimized version, compression may take as much as 6 ms (of course longer than the optimized version, but such execution will affect round trip overall latency, and shorten to maintain 70 to 80 ms (e.g. general The distance allowed through the Internet may be reduced). Other latency may be required). The round trip latency of the Internet 454, user ISP 453, and user premises routing 452 has already been considered, so what remains is video stretch 458 latency, whether video stretch 458 runs on dedicated hardware, or If run in the software of the client device 415 (PC or mobile device), it may be very dependent on the size of the display and the performance of the expanding CPU. Typically, stretch 458 takes between 1 and 8 ms.

Thus, by adding together the worst-case latency seen in the implementation, we can determine the worst-case round-trip latency that can be expected to be experienced by the user of the system shown in FIG. 4A. They are: 1 + 1 + 25 + 22 + 1 + 16 + 6 + 8 = 80ms. And, in practice (with the warnings described below), this uses a prototype version of the system shown in FIG. The round trip latency is shown. Of course, better scenarios than the worst case can result in much shorter latency, but they are not needed to develop widely used commercial services.

In order to achieve the latency listed in FIG. 4B on the general Internet, a video compressor 404 and a video extender 412 in the client 415 are required from FIG. 4A to generate a packet stream of very specific characteristics, and so on. The packet sequence generated over the entire path from the hosting service 210 to the display device 422 is free of delays or excess packet loss, and in particular, through the WAN interface 442 and the firewall / router / NAT 443. Consistently lowers the limits of bandwidth available to users on their Internet connections. Moreover, the video compressor generates a sufficiently powerful packet stream that can withstand the inevitable packet loss and packet reordering that usually occurs in the Internet and network transmissions.

low- Latency  Video Compression ( Low - Latency Video Compression )

To achieve the aforementioned goal, one embodiment leads to a new approach for video compression that reduces peak bandwidth requirements and latency for video to be transmitted. Prior to detailed description of this embodiment, an analysis of current video compression techniques will be provided with respect to FIGS. 5, 6A, and 6B. Of course, if the user is provided with sufficient bandwidth to handle the rate required by this technology, this technology will be adopted in accordance with the essential theory. Note that audio compression is not described herein, except to mention that video compression and synchronization are performed simultaneously. Prior art audio compression techniques exist to meet the requirements for this system.

5 illustrates a particular prior art for video compression in each individual video frame 501-503 compressed by compressor logic 520 using a particular compression algorithm to produce a series of compressed frames 511-513. Doing. One embodiment of this technique is a "motion JPEG" in which each frame based on a discrete cosine transform (DCT) is compressed according to a Joint Picture Expert Group (JPEG) compression algorithm. Various other forms of compression algorithms will be employed, while still conforming to this fundamental theory (eg wavelet-based compression algorithms such as JPEG-2000).

One problem with this type of compression is that it reduces the transmission rate of each frame, but it does not exploit the similarity between successive frames to reduce the transmission rate of the entire video stream. For example, assuming that 604 × 480 × 24 bits / pixel = 604 * 480 * 24/8/1024 = 900 kilobytes / frame (KB / frame), as shown in FIG. 5, for an image of a given quality, a motion JPEG Compresses only at a rate of 10, resulting in a data stream of 90 KB / frame. At 60 frames / second, this would require channel bandwidth of 90 KB * 8 bits * 60 frames / second = 42.2 Mbps, which would be much higher bandwidth than almost all home internet connections in the United States today, and too high for many office internet connections. It will be bandwidth. Indeed, if a constant data stream is required at such a high bandwidth, even in an office LAN environment it will provide only one user, which will consume most of the bandwidth of a 100 Mbps Ethernet LAN and support an Ethernet switch. Will make it much more burdensome. Thus, compression for video is inefficient compared to other compression techniques (as described below). Moreover, single-frame compression algorithms such as JPEG and JPEG-2000 may not be able to recognize still images (e.g. artifacts on dense leaves in the scene will not be represented as artifacts because they do not know exactly how dense the leaves can be). Use a lossy compression algorithm that produces unknown compression artifacts. However, once the scene moves, the artifacts can be noticeable because the eye finds artifacts that change from frame to frame, despite the fact that artifacts in one area of the scene cannot be recognized as still images. This is the result of "background noise" recognition in the sequence of frames, and is similar to that seen in "snow" noise that can be seen during minor analog TV reception. Of course, this type of compression will still be used in some embodiments described herein, but generally speaking, in order to avoid background noise in the scene, a high transmission rate (eg, a low compression ratio) is required for a given perceptual quality.

Other forms of compression, such as H.264 or Windows Media VC9, MPEG2 and MPEG4, are all efficient for compressing video streams because they take advantage of similarity between successive frames. All of these techniques follow the same general technique for compressing video. Thus, although the H.264 standard is described, the same general theory applies to other compression algorithms. A large number of H.264 compressors and expanders are available and include the x264 open source software library for compressing H.264 and the FFmpeg open source software library for extending H.264.

6A and 6B show that a series of uncompressed video frames 501-503, 559-561 are converted into a series of "I frames" 611, 671 by compression logic 620; "P frames" 612, 613; And exemplary conventional compression techniques that are compressed into "B frames " In FIG. 6A the vertical axis generally represents the resulting size of each encoded frame (although the frame is not shown in scale). As described, video coding using I frames, B frames, and P frames will be apparent to those skilled in the art, and I frame 611 is a completely uncompressed frame 501 (as described above). DCT based compression (similar to compressed JPEG images). P frames 612 and 613 are typically significantly smaller than I frames 611. Because they use data in the preceding I frame or P frame; That is, they contain data indicating changes between preceding I or P frames. The B frame 670 is similar to that of the P frame except that the B frame uses the frame for the next reference frame as well as the potential frame in the preceding reference frame.

In the following discussion, the requested frame rate is 60 frames / second, each I frame is approximately 160 Kb, the average P frame and B frame is 16 Kb and a new I frame is generated every second. With this set of parameters, the average bit rate would be: 160Kb + 16Kb * 59 = 1.1 Mbps. This rate falls within the maximum rate of many current broadband Internet connections for home and office. This technique also tends to avoid background noise problems only from intraframe encoding because P frames and B frames track differences between frames, so that compression artifacts reduce or reduce the background noise problem described above. It tends to disappear and not appear in two frames.

One problem with the aforementioned type of compression is that even if the average rate is generally low (eg 1.1 Mbps), a single I frame will take several frame times to transmit. For example, a 2.2 Mbps network connection using a prior art (e.g., a DSL or cable modem with a 2.2 Mbps peak at the maximum available rate 302 from Figure 3A) is typically at 1.1 Mbps with 160 Kbps I frames, each 60 frames. It would be appropriate to stream the video. This can be accomplished by having an expander that queues up one second of video before stretching the video. In one second, 1.1 Mb of data will be transmitted, which will easily be accommodated up to 2.2 Mbps, the maximum available rate, and the available rate will drop periodically to over 50%. Unfortunately, this prior art approach will result in one second latency for the video because of the one second video buffer in the receiver. Such a delay is suitable for many leading applications (eg, playback of linear video), but it is too long a latency for fast motion video games that cannot tolerate more than 70-80 ms latency.

If there is an attempt to reduce the one second video buffer, it will still not result in proper omissions in latency for fast motion video games. In some cases, as described above, the use of B frames will require the reception of not only I frames, but also all B frames that precede the I frames. If we assume 59 non-I frames divided roughly between P and B frames, there will be at least 29 B frames and one I frame received before any B frames are displayed. Thus, despite the available channel bandwidth, it will require a delay of 29 + 1 = 30 frames or a latency of 500 ms in each 1/60 second period. Obviously it is too long.

Thus, the other approach will eliminate B frames and use only I frames and P frames. (One result of this will be increased bit rate for a given quality level, but in this example each I frame is 160Kb for consistency. And the average P frame is 16Kb in size, so the rate still continues to assume 1.1 Mbps). Remove it. One problem with this approach is that, as is typical in most homes and many offices, I-frames on low-bandwidth channels are too large for average P frames, and the transmission of I-frames substantially adds latency. This is illustrated in Figure 6b. The video stream rate 624 is below the maximum rate 621 available except in the case of I frames, where the peak rate required for the I frame 623 is the maximum rate 622 available (and rated). Even the maximum data rate 621). The rate required by the P frame is less than the maximum rate available. Although the maximum rate peak available at 2.2 Mbps remains steady at its 2.2 Mbps peak rate, it will take 160Kb / 2.2Mb = 71 ms to transmit an I frame, and if the maximum rate 622 available is up to 50% ( 1.1 Mbps), it will take 142 ms to transmit an I frame. So, the latency for sending I frames will fall somewhat between 71 and 142 ms. This latency is added to this latency added to the latency identified in FIG. 6B, in the worst case up to 70 ms, which is until the image appears on the display device 422 at the time when the user operates the input device 421. This will result in a total round trip latency of 222 ms, which is too high. And if the maximum available rate falls below 2.2 Mbps, the latency will be further increased.

There is also "jamming" which is a serious consequence for ISPs with peak rates 623 that are generally well above the available rates 622. The mechanisms at different ISPs will behave differently, but the following behaviors are quite common between DSL and cable modem ISPs when receiving packets at rates much higher than the available rates 622: (a) By queuing them, Delay (induce latency), (b) drop some or all packets, and (c) deactivate the connection for a period of time (most likely because of the malicious attack, such as a "denial of service"). Thus, transmitting a packet stream at the maximum rate with the characteristics as shown in FIG. 6B is not a viable option. Peak 623 will be queued up at hosting service 210 and transmits at a rate lower than the maximum available rate leading to the unacceptable latency described in the preceding paragraph.

Moreover, the video stream rate sequence 624 shown in FIG. 6B is a well "tame" video stream rate sequence and is expected to result from compressing the video into a video sequence that has very little motion and does not change very much. It will be a rate sequence of classifications (e.g., the camera is in a fixed position, little motion, and will be common in video conferencing).

The video stream rate sequence 634 shown in FIG. 6C is a typical sequence found in motion pictures or video with much more movement made in a video game or some application software. Note the addition to I frame peak 633, and there are also P frame peaks (such as 635 and 636) that exceed the maximum rate available in many cases and are quite large. Although these P frame peaks are not quite as large as the I frame peaks, they are still too large to be carried by the channel at the maximum data rate, and like the I frame peaks, the P frame peaks must be transmitted slowly (thereby increasing latency).

In a high bandwidth channel (eg, 100 Mbps LAN, or high bandwidth 100 Mbps private connection) the network will be able to tolerate large peaks, such as I frame peak 633 or P frame peak 636, and theoretically low-latency will be maintained. I can. However, such networks are frequently shared among many users (e.g. in an office environment), and such "peaky" data, especially if network traffic is transmitted over private shared connections (e.g., from data centers to offices). Will affect the performance of the LAN. First of all, it should be noted that this example is typically a low resolution video stream of 640x480 pixels at 60fps. 1920 × 1080 of HDTV streams at 60fps are easily handled by modern computers and displays, and 2560 × 1440 resolution displays at 60fps are increasingly available (eg, Apple, Inc's 30 inch displays). A high motion video sequence of 1920x1080 at 60fps will require 4.5Mbps using H.264 compression for reasonable quality levels. If we assume an I frame peak at 10 × nominal rate, it will result in a 45Mbps peak, but it is not only small but still a significant P frame peak. If some users receive video streams on the same 100 Mbps network (for example, a private network connection between the office and the data center), how can you adjust the peaks from some of the video streams, overwhelm the network's bandwidth, It is easy to know if the bandwidth of the backplate of the supporting switch is potentially overwhelming. Even in a gigabit Ethernet network, if enough users are fully tuned at once, it can overcome a network or network switch. And, once 2560x1440 resolution video becomes more common, the average video stream rate will be 9.5 Mbps, possibly resulting in a 95 Mbps peak rate. Needless to say, the 100 Mbps connection between the data center and the office (which is an exceptionally fast connection today) will be flooded by peak traffic from a completely single user. Thus, although LAN and private network connections can tolerate picky streaming video better, streaming video with high peaks is undesirable and will require special planning and facilities by the IT department of the office.

Of course, for standard linear video applications this issue is not a problem. Because the rate is " smoothed " at the point of data and transmission for each frame below the maximum available rate 622, the buffer at the client stores a sequence of I, P, B frames before it is stretched. Thus, the rate on the network remains close to the average rate of the video stream. Unfortunately, this leads to latency even though B frames are not used, which is not acceptable for low-latency applications such as video games and applications that require fast response times. A prior art solution to mitigate high peak video streams is to use an encoding technique called "Constant Bit Rate (CBR)". Although the term CBR seems to mean that all frames are compressed to have the same bit rate (eg, size), in general it refers to a certain number of frames (in our case, one frame). Compression paradigm that allows the maximum bit rate across. For example, in the case of FIG. 6C, if the CBR constraint is applied to encode a limited bit rate, for example a rated maximum data rate 621 of 70%, then the compression algorithm will limit the compression of each frame so that the maximum rated The rate 621 will be compressed to fewer bits. The result of this is that frames that generally require more bits to maintain a given quality level are “starved” to bits and the image quality of the frames requires more bits than 70% of the maximum bit rate 621. It will be worse than other frames that don't. This approach can produce acceptable results for some form of compressed video where (a) small motion or scene changes are expected and (b) the user can periodically accept quality reduction. A good example of an application that goes well with CBR is video teleconferencing because there are few peaks, and if the quality is clearly reduced (e.g. if it's shot on a camera, while it's being shot without enough bit for high quality image compression), Can result in reduced image quality), which is mostly acceptable to the user. Unfortunately, CBR is not well suited for many other applications that have high complexity scenes or require a lot of movement and / or a reasonable level of quality.

The low-latency compression logic 404 employed in one embodiment uses several different techniques for streaming the low-latency compressed video and delivering it to a range of problems. First, low-latency compression logic 404 generates only I frames and P frames, thereby mitigating the need to wait several frame times to decode each B frame. Furthermore, in one embodiment, as shown in FIG. 7A, the low-latency compression logic 404 stores each uncompressed frame 701-760 as a series of "tiles" and individually I frames or P frames. Encode into each tile such as The group of compressed I-frames and P-frames is referred to herein as "R frames" 711-770. In the specific example shown in FIG. 7A, each uncompressed frame is subdivided into a 4 × 4 matrix of 16 tiles. However, this fundamental principle is not limited to any particular subdivision plan.

In one embodiment, low-latency compression logic 404 divides the video frame into multiple tiles, and an I frame (e.g., the tile is as if it is a discrete video frame of 1/16 ^th to the size of the entire image, such as The compression used for "mini" frames is I frame compression, which encodes (eg, compresses) one tile from each frame, and P frames (eg, each "mini" 1/16 ^th frame The compression used for this is P frame compression). Tiles compressed as I frames and P frames will generally be referred to as "I tiles" and "P tiles". With each successive video frame, the tile encoded as an I tile changes. Thus, at a given frame time, only one of the tiles in the video frame is an I tile, and the rest of the tiles are P tiles. For example, in FIG. 7A, tile 0 of uncompressed frame 701 is encoded as I tile I ₀ and the remaining 1-15 tiles are encoded from P tile P ₁ to P ₁₅ to produce R frame 711. do. In the next uncompressed video frame 702, tile 1 of uncompressed frame 701 is encoded as I tile I ₁ and the remaining tiles 0 and 15 in tile 2 are P files to produce an R frame 712. It is encoded as P ₁₅ at P _{0 ,} P ₂ . Thus, for tiles, the I and P tiles are interleaved in time over successive frames. The process continues until the R tile 770 generates the last tile in the matrix encoded as an I tile (eg, I ₁₅ ). The process then resumes generating another R frame like frame 711 (e.g., encoding an I tile for tile 0). Although not shown in FIG. 7A, in one embodiment, the first R frame of the video sequence of the R frame includes only I tiles (eg, a subsequent P frame has reference frame data therefrom to associate with operation). do. Alternatively, in one embodiment, the startup sequence generally uses the same I tile pattern but does not include P tiles for tiles that have not yet been encoded as I tiles. In other words, some tiles are not encoded with any data until the first I tile arrives, thereby avoiding the startup peak at video stream rate 934 in FIG. 9A, which is described in detail below. Moreover, other sizes and shapes may be used for the tiles, as described below, while adhering to this basic theory.

Video stretching logic 412 operating at client 415 stretches each tile as if it were a separate video sequence of I and P frames, and then renders each tile for the frame buffer driving display device 422. . For example, I ₀ and P ₀ from R frame 711 to frame 770 are used to reconstruct tile 1 and the like. As mentioned above, the stretching of I frames and P frames is well known in the art, and the stretching of I tiles and P tiles can be achieved by having multiple instances of video stretching running on the client 415. . Although the multi-operational process seems to increase the computer burden on the client 415, it is not really because the tiles themselves are proportionally smaller with respect to many additional processing, so the number of pixels displayed is one processing. If yes, it is the same as using conventional full size I and P frames.

This R frame technique significantly mitigates the typical bandwidth peak associated with the I frames shown in FIGS. 6B and 6C because some given frames typically consist of P frames that are typically smaller than I frames. For example, assuming a typical I frame is 160 Kb, the I tiles of each of the frames shown in FIG. 7A will be approximately 1/16 or 10 Kb of this size. Similarly, assuming a typical P frame is 16 Kb, the P frame for each of the tiles shown in FIG. 7A will be approximately 1 Kb. The final result of the R frame is approximately 10Kb + 15 * 1Kb = 25Kb. Thus, each 60-frame sequence is 25 Kb * 60 = 1.5 Mbps. So at 60 frames / second, this would require channel performance capable of maintaining a bandwidth of 1.5 Mbps, but would require much lower peaks due to the I tiles distributed throughout the 60-frame interval.

Note the previous example with the same assumed rate for I frames and P frames, the average rate is 1.1 Mbps. This is because in the previous example, a new I frame was introduced only once every 60 frame times, whereas in this example, the 16 tiles that make up the I frame cycle are completed in 16 frame times, so this equivalent of I frames. Is derived every 16 frame times, resulting in a slightly higher average rate. Derivation of more frequent I frames does not linearly increase the rate. This is due to the fact that a P frame (or P tile) encodes the difference between the next frame in the preceding frame. So, if the preceding frame is significantly different from the next frame, the P frame will be very large. However, since the P frame is usually derived from the preceding frame rather than the actual frame, the encoded frame result will contain more errors (eg, visual artifacts) than I frames with the appropriate number of bits. And, when one P frame follows another P frame, what can happen is an accumulation of errors that gets worse when there is a long sequence P frame. Now, a sophisticated video compressor will find that the quality of the image is degraded after P frames of the sequence, and if necessary, it will allocate more bits to subsequent P frames to improve the quality, but if it is the most efficient This is the course of the course and replaces P frames with I frames. So, when long sequences of P frames are used (e.g. 59 frames, as in the previous example), especially when the scene has a lot of complexity and / or motion, typically as more bits are removed from the I frame, Needed for P frames.

Or, to see a P frame at an opposite point in time, P frames that closely follow the I frame tend to require fewer bits than P frames that are further removed from the I frame. Thus, in the example shown in FIG. 7A, the P frame is not added more than 15 frames removed from the I frame, and as in the previous example, the P frame may be 59 frames removed from the I frame preceding it. Therefore, as I frames are more frequent, P frames are smaller. Of course, the exact related size will vary based on the characteristics of the video stream, but in the example of FIG. 7A if an I tile is 10Kb, usually a 0.75Kb size P tile, 10Kb + 15 * 0.75Kb = 21.25Kb, or At 60 frames per second, the rate would be 21.25 Kb * 60 = 1.3 Mbps, or about 16% higher than the stream with I frames followed by 59 P frames at 1.1 Mbps. Once again, the relevant outcomes between these two approaches to video compression will vary depending on the video sequence, but typically we use R-frames at a given level of quality than using I / P frame sequences. Empirically we find that it requires about 20% more bits for However, of course R frames dramatically reduce the peaks that make video sequences usable with much less latency than I / P frame sequences.

The R frame can be configured in a variety of different ways depending on the nature of the video sequence, channel reliability, and available bit rate. In an alternative embodiment, a large number of other tiles are used over 16 in a 4x4 configuration. For example, 2 tiles would be used in 2 × 1 or 1 × 2, 4 tiles would be used in a 2 × 2, 4 × 1, or 1 × 4 configuration, or 6 tiles would be 3 × 2, 2 × 3, 6 × 1 , Or in an 1 × 6 configuration, or 8 tiles may be used in a 4 × 2 (as shown in FIG. 7B), 2 × 4, 8 × 1, or 1 × 8 configuration. The tiles need not be square, and the video frame need not be square or even rectangular. Tiles can be split into whatever form is appropriate for use in video streams and applications.

In another embodiment, the cycling of the I and P tiles is not fixed to multiple tiles. For example, in a 4 × 2 configuration 8 tile, a 16 cycle sequence can be used as shown in FIG. 7B. Sequentially uncompressed frames 721, 722, 723 are divided into 8 tiles, 0-7 each, and each tile is compressed separately. In the R frame 731, only 0 tiles are compressed as I tiles and the remaining tiles are compressed into P tiles. During the next R frame 732 all 8 tiles are compressed as P tiles, then during the next R frame 733 one tile is compressed as an I tile and all other tiles are compressed as P tiles. Then, the sequence is continued for 16 frames, with I frames created with only all other frames, so the last I tile is generated for 7 tiles for 15 ^th frame time (not shown in FIG. 7B) and for 16 ^th frame time. R frame 780 is compressed using all P tiles. Then, the sequence starts again with zero tiles compressed as I tiles and the other tiles are compressed as P tiles. As in the previous embodiment, the entire video sequence of the first frame will typically be all I tiles to provide a reference for the P tiles at that point. Cycling I and P tiles does not require multiple tiles. For example, with 8 tiles, each frame with an I tile may follow up to 2 frames with all P tiles until another I tile is used. In yet another embodiment, some tiles will be sequenced into I tiles more frequently than other tiles. For example, some areas of the screen are known to have more motion than is required from frequent I tiles, while other areas are more frequent. Less static required from I tiles (eg, showing scores for a game). Moreover, although each frame is shown as a single I tile in Figures 7A and 7B, multiple I tiles will be encoded in a single frame (depending on the bandwidth of the transport channel). Conversely, some frames or frame sequences may be sent without having I tiles (eg, only P tiles).

A good reason for the preceding paragraph study approach is that it does not seem to have I tiles distributed over every single frame, resulting in larger peaks, and the operation of the system is not simple. Each tile is compressed separately from other tiles, and tiles can be less efficient as the encoding of each tile is smaller because the compressor of a given tile cannot use similar image features and similar behavior from other tiles. Thus, dividing the screen into 16 tiles will generally be less efficient than dividing the screen into 8 tiles. However, if the screen is 8 tiles it causes the data of the entire I frame to be derived every 8 frames instead of every 16 frames, which results in a very high bit rate overall. Thus, by deriving the entire I frame every 16 frames instead of every 8 frames, the overall rate is reduced. In addition, by using eight larger tiles instead of 16 smaller tiles, the overall data rate will be reduced, which will also mitigate to some extent the data peaks caused by the large tiles.

In another embodiment, the low-latency video compression logic 404 in FIGS. 7A and 7B is preconfigured by a setting based on a known feature of the video sequence to be compressed, or automatically based on analysis of image quality in each tile in progress. To control the allocation of bits for various tiles in an R frame. For example, in some racing video games, the player's front of the car (which usually has no motion in the scene) occupies a large area in the bottom half of the screen, while the top half of the screen is filled with roads, buildings, and landscapes that are entirely accessible. This is almost always moving. If the compression logic 404 allocates the same number of bits to each tile, then the tiles (tiles 4-7) at the bottom half of the screen in the uncompressed frame 721 in FIG. 7B, generally in FIG. 7B. In the uncompressed frame 721 will be compressed to a higher quality than the tiles (tiles 0-3) in the top half of the screen. If this particular game, or particular scene of the game, is known to have such a feature, the operator of the hosting service 210 may use compression logic to allocate more bits for the tile at the top of the screen than the tile at the bottom of the screen. 404). Or, the compression logic 404 can measure the quality of the tile's compression after the frame is compressed (using one or more compression quality performance indicators, such as Peak-To-Noise Ratio (PSNR)), If a decision is made over the time of a particular window, some tiles will consistently produce better quality results, and gradually until the various tiles reach a similar level of quality, it will gradually produce more bits with lower quality results. Assign to In an alternative embodiment, the compressor logic 404 assigns bits to achieve high quality for a particular tile or group of tiles. For example, it will provide a better perceptual appearance overall to have a higher quality at the center of the screen than the edges.

In one embodiment, to improve the resolution of certain areas of the video stream, video compression logic 404 generally has more complex scenes and / or motions than areas of the video stream with fewer complex scenes and / or motions. Use smaller tiles to encode regions of the video stream. For example, as shown in FIG. 8, smaller tiles are adopted around a moving character 805 in an area of the R frame 811 (potentially with the same tile size and following a series of R frames (not shown). Not)). When the character 805 then moves to a new area of the image, a smaller tile is used around this new area in another R frame 812 as shown. As mentioned above, while still following this fundamental theory, various other sizes and shapes will be employed as "tiles."

While the periodic I / P tiles described above essentially reduce the peaks in the bitrate of the video stream, they are particularly fast changing or very complex video images, such as those generated by motion pictures, video games, and any application software. In the case of, the peak is not removed entirely. For example, during a sudden scene change, a complex frame will be followed by another completely complex frame. Although some I tiles may be before scene transitions up to only a few frame times, they cannot help in this situation because the material of the new frame is not related to the pre-I tile. In such situations (and in many other situations where many images change, although not all), video compressor 404 can determine how much if not all of the P tiles coded more efficiently, such as I tiles, and what consequences It will determine if it is a very large peak in the bit rate for that frame.

As discussed earlier, most consumer-grade Internet connections (and many office connections) are simple, and they simply exceed the maximum available data rate shown as 622 in FIG. 6C, according to the maximum rated data rate 621. (jam) "Not possible with data. The rated maximum transfer rate 621 (eg, “6 Mbps DSL”) is essentially a marketing number for a user considering purchasing an internet connection, but in general it does not guarantee any level of performance. For the purpose of this application, it is meaningless, and our only concern is the maximum rate 622 available at the time the video is streamed over the connection. As a result, in Figures 9A and 9C, when we describe a solution for the peaking problem, the rated maximum data rate is omitted from the graph, and only the maximum data rate 922 available is shown. The video stream rate should not exceed the maximum available rate 922.

To handle this, the first thing that video compressor 404 has to do is determine the peak rate 941, which is the rate the channel can handle continuously. This speed is determined by many techniques. One such technique would be to gradually send an increasing high rate test stream from the hosting service 210 to the client 415 in FIGS. 4A and 4B, providing the client with feedback about some level of packet loss and latency to the hosting service. something to do. When packet loss and / or latency starts to show a sharp increase, it is an indication that the maximum rate 922 available is reaching. Thereafter, the hosting service 210 may gradually reduce the transmission rate of the test stream until the client 415 receives a minimum amount of packet loss and latency that the test stream allows for a reasonable period of time. Over time, the peak rate 941 will fluctuate (e.g., if another user in the home uses the Internet connection excessively), the client 415 will determine whether packet loss or latency has been increased, and the maximum rate available 922. ) Needs to be constantly monitored to see if it falls below a pre-established peak rate 941, and if so is a peak rate 941. Similarly, if time passes and the client 415 finds that packet loss and latency remain at an optimal level, it may request that the video compressor slowly increase the rate to see if the maximum rate available is increased ( For example, if another user in the home stops using an excessive Internet connection), wait again until packet loss and / or high latency indicate that the available transfer rate 922 has been exceeded, again at a lower transfer rate (941). But one is probably higher than that level before testing the increased bit rate. Thus, by using this technique (and other such techniques), the peak rate 941 can be found and adjusted periodically as needed. Peak rate 941 will be set to the maximum rate that can be used by video compressor 404 to stream video to the user. Logic for determining the peak rate may be executed at the user premises 211 and / or hosting service 210. In the user premises 211, the client 415 performs an operation for determining the peak transfer rate and sends this information back to the hosting service 210; In the hosting service 210, the server 402 in the hosting service performs operations to determine the peak rate based on statistics received from the client 415 (eg, packet loss, latency, maximum rate, etc.).

9A illustrates an example of a video stream rate 934 having substantially complex scenes and / or motions generated using the recursive I / P tile compression technique shown in FIGS. 7A, 7B, and 8 and described previously. It is shown. The video compressor 404 is configured to output compressed video at an average rate below the peak rate 941, and most of the time, the video stream data stays below the peak rate 941. Comparison of the rate 934 and the video stream rate 634 shown in FIG. 6C made using I / P / B or I / P frames shows that cyclic I / P tile compression produces a smoother rate. Still, frame 2X peak 952 (which is close to 2X peak rate 942) and frame 4X peak 954 (which is close to 4X peak rate 944), whose rate exceeds peak rate 941. This cannot be tolerated. Indeed, from fast-changing video games to high video, peaks above the peak rate 941 occur less than 2% of the frames, peaks above the 2X peak rate 942 rarely occur, and 3X peak rate 943 Peaks exceeding rarely occur. However, when they happen (eg during scene change), the bit rate required by them is necessary to produce a good quality video image.

One way to solve this problem is to install a video compressor 404 so that its maximum rate output is peak rate 941. Unfortunately, the resulting video output quality during the peak frame is poor because the compression algorithm is "lack" for the bits. As a result of the appearance of compression artifacts when there is a sudden transition or rapid movement, then the user notices artifacts that always occur suddenly when there is a sudden change or rapid movement, and they can become quite annoying.

Although the human visual system is quite sensitive to visual artifacts that appear during sudden changes or rapid movements, it is not very sensitive to detecting a decrease in the speed of the frame in such a situation. In fact, when such a sudden change occurs, it happens ahead of time as the human visual system tracks the change, which is not noticed if the frame rate obviously drops from 60fps to 30fps and then immediately returns to 60fps. And, in the case of very dramatic transformations such as sudden scene changes, the human visual system does not notice if the frame rate drops to 20 or 15 fps, and then immediately returns to 60 fps. While frame rate changes only occur infrequently, to human observers, it indicates that the video is running continuously at 60fps.

This feature of the human visual system is utilized by the technique shown in FIG. 9B. Server 402 (in FIGS. 4A and 4B) outputs an uncompressed video output stream at a steady frame rate (60 fps in one embodiment). The timeline for each frame (961 - 970) show the 1/60 ^th of a second output. Starting with frame 961, each uncompressed video frame compresses the frame in less than one frame time to produce compressed frame 1 981 for the first frame. Is output to 404. The data produced for compressed frame 1 981 may be larger or smaller depending on many factors, as described above. If the data is small enough, it can be sent to the client 415 in one frame time (1/60 ^th seconds) or less than the peak rate 941, followed by the transmission time (Xmit time) 991 (arrow). Length indicates the duration of the transmission time). At the next frame time, server 402 produces frame 2 962, which is not compressed, which is compressed into frame 2 982, which is sent to client 415 during transfer time 992, which is It is less than the frame time of the peak data rate (941).

Next, at the next frame time, server 402 generates frame 3 963, which is not compressed. When it is compressed by video compressor 404, the resulting compressed frame 3 983 is larger data than can be transmitted at peak rate 941 at one frame time. Thus, it is transmitted during transmission time (2X peak) 993, which occupies all frame time and a portion of the next frame time. Now, during the next frame time, the server 402 creates another uncompressed frame 4 964 and outputs it to the video compressor 404 but the data is ignored and illustrated at 974. This is because the video compressor 404 is configured to ignore the added uncompressed video frame that has arrived while still transmitting the previous compressed frame. Of course the video extender of client 415 fails to receive frame 4, but it simply continues to display frame 3 for two frame times on display device 422 (e.g., the frame rate is briefly reduced from 60fps to 30fps). Is reduced).

For the next frame 5, server 402 outputs uncompressed frame 5 965, which is compressed into compressed frame 5 985, and is transmitted within one frame during transmission time 995. The video extender of client 415 expands frame 5 and displays it on display device 422. Next, server 402 outputs uncompressed frame 6 966, which is compressed into frame 6 986, which is compressed by video compressor 404, but at this time the resulting data is very large. The compressed frame will be transmitted during transmission time (4X peak) 996 at peak rate 941, but it takes almost four frame times to transmit the frame. During the next three frame times, video compressor 404 ignores three frames from server 402, and the extender of client 415 maintains frame 6 on display device 422 for four frames of time (e.g., Obviously reduce the frame rate from 60fps to 15fps). Finally, server 402 outputs frame 10 970, and video compressor 404 compresses it into compressed frame 10 987, which is transmitted during transmission time 997, and client 415. ) Stretches frame 10 and displays it on display 422 and once again the video starts again at 60 fps.

Although the video compressor 404 drops from the video stream generated by the server 402 to a video frame, no matter what type of incoming audio it is, it does not drop audio data, the video frame falls and it is sent to the client 415. When transmitted, it continues to compress the audio data, which continues to decompress the audio data and provides the audio to any device used by the user to play the audio. Therefore, the audio continues without dropping even during the period in which the frame falls. Compared with compressed video, compressed audio typically consumes a small percentage of bandwidth and consequently has no major impact on the overall rate. Although it is not shown in any rate diagram, there is always a limited rate capability for compressed audio streams within peak rate 941.

The example shown in FIG. 9B is selected to show how much the frame rate falls during the rate peak, but video games, movies, and certain applications, such as the rate peak, when cyclic I / P tile techniques are used as described before Frames that drop continuously are rare, even in high complex scenes / high motion sequences such as those that occur in software. As a result, reduced frame rates are not frequent and obviously the human visual system does not detect them.

If the frame rate reduction mechanism is described to apply to the video stream rate described in FIG. 9A, the resulting video stream rate is shown in FIG. 9C. In this example, the 2x peak 952 is reduced to the flattened 2x peak 953 and the 4x peak 955 is reduced to the flattened 4x peak 955 such that the entire video stream data 934 is peak rate 941. ) Or remain below it.

Thus, using the techniques described above, high-motion video streams can be transmitted over common Internet and consumer-grade Internet connections with low latency. Moreover, in office environments, LANs (eg, 100 Mbs Ethernet or 802.11g wireless) or personal networks (eg, 100 Mbps connections between data centers and offices), high-motion video streams can be transmitted without peaks, allowing multiple users (eg, Sending 1920 × 1080 at 4.5 Mbps to 60 fps) can use an overwhelming network or network switch backplate, shared personal data connection without overlapping peaks

Rate adjustment ( Data Rate Adjustment )

In one embodiment, hosting service 210 evaluates the maximum available rate 622 and the latency of the channel and dynamically adjusts the rate in response to determining the appropriate rate for the first video stream. To adjust the rate, the hosting service 210 may modify the image resolution and / or multiple frames / second of the video stream sent to the client 415, for example. The hosting service can also adjust the quality level of the compressed video. Changing the resolution of the video stream is, for example, 640 × 360 at 1280 × 720 resolution, so that video stretching logic 412 on the client 415 can scale the image proportionally to maintain the same image size on the display screen. have.

In one embodiment, in some situations where the channel is completely out of service, hosting service 210 stops the game. In the case of a multiplayer game, the hosting service reports to the other user whether the user left the game and / or stopped the game for another user.

Dropped or delayed packets ( Dropped or Delayed Packets )

In one embodiment, packets received in the order of instructions that arrive too late for stretching due to packet loss or latency requirements of uncompressed frames between video compressor 404 and client 415 in FIG. 4A or 4B. If data is lost due to the video stretching logic 412 may mitigate visual artifacts. In I / P frame execution streaming, if there are lost / delayed packets, the entire screen is affected, potentially causing the screen to freeze for a period of time or show other screen-wide visual artifacts. For example, if a lost / delayed packet is responsible for the loss of an I frame, then the extender will lack the criteria for every P frame to follow until a new I frame is received. If a P frame is lost, then it will affect the P frame for the entire screen following it. Depending on how long it will take before an I frame appears, it will have a longer or shorter visual impact. Using interleaved I / P tiles as shown in FIGS. 7A and 7B, lost / delayed packets have much less impact on the entire screen because it will only affect the tiles included in the affected packet. If the data of each tile is sent in a separate packet, then if the packet is lost, it will only affect one tile. Of course, the duration of the visual artifact will depend on whether the I tile packet is lost, and if the P tile is lost, how many frames will take until the I tile appears. However, what is given to another tile on the screen is very frequently updated to I frames (potentially every frame), and may not be the other tile even when one tile is affected on the screen. Moreover, if an event causes the loss of several packets immediately (e.g. spike on power next to the DSL line which obviously interrupts the data flow), then some tiles will have more impact than others. However, because some tiles will be updated quickly with new I tiles, they will only be affected briefly. Also, with streaming I / P frame execution, I frames are the most critical frames, but I frames are very large, so if there are events that result in dropped / delayed packets, perhaps I frames rather than smaller I tiles are affected. (Eg, if any part of an I frame is lost, it cannot be stretched at all). For all these reasons, using I / P tiles results in much less visual artifacts than when packets dropped / delayed into I / P frames.

One embodiment attempts to reduce the effects of lost packets by intelligently packaging compressed tiles in a transmission control protocol (TCP) packet or a user datagram protocol (UDP) packet. For example, in one embodiment, tiles are aligned with packet boundaries whenever possible. 10A illustrates how a tile will be packed in a series of packets 1001-1005 without implementing this feature. In particular, in FIG. 10A, the tiles cross the packet boundary and are inefficiently packed so that the loss of a single packet results in the loss of multiple frames. For example, if packet 1003 or 1004 is lost, three tiles are lost, which results in visual artifacts.

In contrast, FIG. 10B illustrates tile packing logic 1010 for intelligently including tiles in a packet to reduce the effects of packet loss. First, tile packing logic 1010 aligns tiles with packet boundaries. Thus, tiles T1, T3, T4, T7 and T2 are generally aligned at the boundary of packets 1001-1005. The tile packing logic also attempts to fix the tile in the packet in the most efficient way possible, without crossing the packet boundary. Based on the size of each tile, tiles T1 and T6 are combined into one packet 1001; Tiles T3 and T5 are combined into one packet 1002; Tiles T4 and T8 are combined into one packet 1003; Tile T8 is added to packet 1004; Tile T2 is then added to packet 1005. Thus, under this structure, a single packet loss will not result in more than 2 tiles loss (rather than 3 tiles as shown in FIG. 10A).

An additional advantage to the embodiment shown in FIG. 10B is that the tiles are sent in a different order than they are displayed in the image. This way, if adjacent packets are lost due to the same event interfering with it being transmitted, it will affect areas that are not close to each other on the screen, creating almost unknown artifacts on the display.

One embodiment employs forward error correction (FEC) technology to protect certain portions of the video stream from channel errors. As is known in the art, FEC techniques such as Reed-Solomon and Viterbi generate and append error correction data information for data transmitted over a communication channel. If an error occurs in the underlying data (eg I frame), then the FEC will then correct the error.

The FEC code causes the transmission rate of the transmission to increase; So ideally, they are used only where they are needed most. If data is sent so that it does not result in highly recognizable visual artifacts, it would be desirable not to use FEC codes to protect the data. For example P tile is immediately ahead of an I tile generated only the artifacts (e. G., Will not be updated in the tiles of the screen) of the time during the 1/60 ^th second from the screen in the case where it is lost. Artifacts of such vision are hardly perceived by the human eye. As the P tile is later than the I tile, losing the P tile becomes progressively more pronounced. For example, if the tile cycle pattern is followed by an I tile up to 15 P tiles before the I tile is available again, then immediately if the P tile following the I tile is lost, it will be for 15 frame times (at 60 fps it is 250 ms). Will result in tiles showing the wrong image. The human eye will be able to easily detect confusion in the stream for 250 ms. Thus, the closer the P tile is from the new I tile (eg, the closer the P tile follows the I tile), the better the artifacts can be recognized. As discussed above, in general, the closer the P tile follows the I tile, the smaller the data for the P tile. Thus, the P tiles following the I tiles are not more critical to protect them from being lost, and they are smaller in size. And in general, the smaller the data that needs to be protected, the smaller the FEC code is required to protect it.

Thus, as shown in FIG. 11A, because of the importance of I tiles in the video stream of one embodiment, only I tiles are provided with FEC codes. Accordingly, FEC 1101 includes an error correction code for I tile 1100 and FEC 1104 includes an error correction code for I tile 1103. In this embodiment, there is no FEC generated for the P tile.

In one embodiment shown in FIG. 11B, the FEC code is also generated for P tiles that are prone to artifacts of time if lost. In this embodiment, the FEC 1105 provides an error correction code for the first three tiles, but not for the subsequent P tiles. In another embodiment, the FEC code is generated for the P tile with the smallest data size (which tends to self-select the P tile that occurs most immediately after the I tile, and is most critical for protection). ).

In another embodiment, tiles are sent twice in separate times in different packets than in sending some tiles and FEC codes. If one packet is lost / delayed, another packet is used.

In one embodiment shown in FIG. 11C, the FEC codes 1111 and 1113 generate audio packets 1110 and 1112 which are generally transmitted from the hosting service simultaneously with the video. It is particularly important to maintain the integrity of the audio in the video stream because distorted audio (eg, clicking or hissing) will result in a particularly unpleasant user experience. The FEC code helps to ensure that audio content is rendered at the client computer 415 without distortion.

In another embodiment, rather than sending audio data and FEC codes, audio data is sent twice each time in a different packet. If one packet is lost / delayed, another packet is used.

Additionally, in one embodiment shown in FIG. 11D, the FEC codes 1121 and 1123 are generally (eg, button pressed) user input instructions 1120 and upstream transmitted from the client 415 to the hosting service 210. 1122). This is important because missing a button press or mouse movement in a video game or application can result in an unpleasant experience for the user.

In another embodiment, rather than transmitting the user input command data and the FEC code, the user input command data is sent twice, each time in a different packet. If one packet is lost / delayed, another packet is used.

In one embodiment, the hosting service 210 evaluates the quality of the communication channel with the client 415 to determine whether to use the FEC code, and if using the FEC code, where the FEC will be applied to video, audio, and the user. Evaluate what part of the command it is. The "quality" assessment of the channel will include such functions as measuring packet loss, latency, etc. as described above. If the channel is not particularly reliable then hosting service 210 will apply FEC to all I tiles, P tiles, audio and user instructions. In contrast, if the channel is reliable, then the hosting service 210 will only apply FEC to audio and user commands, or not to apply FEC to audio or video, or not to use FEC at all. Many other changes in the application of FEC could still be adopted following this fundamental theory. In one embodiment, the hosting service 210 continuously monitors the state of the channel and changes in FEC policy accordingly.

In another embodiment, referring to FIGS. 4A and 4B, when a packet is lost / delayed, the FEC may not be able to compensate for the tile data loss, possibly resulting in loss of tile data, or perhaps because of certain bad packet loss, and the client 415. ) Evaluates how many frames were left before the new I tile was received and compares round trip latency from client 415 to hosting service 210. If the round trip latency is less than a number of frames before the new I tile is due to arrive, then the client 415 sends a message to the hosting service 210 to request the new I tile. This message is routed to the video compressor 404 and generates an I tile, rather than generating a P tile while the data of the tile is lost. The system shown in Figures 4A and 4B is typically designed to provide a round trip latency of less than 80 ms, which is 80 ms (frame at 60 fps has a duration of 16.67 ms, so the 80 ms latency at full frame time is corrected within 83.33 ms). Will result in a tile, which is much less noticeable than 5 frame time-recognizable confusion, but for example 250 ms confusion for 15 frames). Compressor 404 generates I tiles outside its usual cyclic order, and if I tile results in a frame bandwidth that exceeds the available bandwidth, compressor 404 will delay the cycle of other tiles Other tiles will receive the P tile during the frame time (although one tile will normally receive an I tile during that frame), starting with the next frame will normally continue cycling, and normally receive an I tile in the preceding frame. The tile to be played will receive an I tile. Although this operation simply delays the state of R frame cycling, it may not be recognized as normal time.

Video and audio compressors / Elongation  Execution( Video and Audio  Compressor / Decompressor Implementation )

FIG. 12 illustrates a particular embodiment in which multi-core and / or multi-processor 1200 are used to compress 8 tiles in parallel. In one embodiment, a dual processor, quad core Xeon CPU computer system operating at 2.66 GHz or higher, with each core running an open source x264 H.264 compressor as an independent processor, is used. However, various other hardware / software configurations will be used while following this fundamental theory. For example, each CPU core can be replaced with an H.264 compressor running on an FPGA. In the example shown in FIG. 12, cores 1201-1208 are used to process I and P tiles simultaneously as eight independent threads. As is well known in the art, current multi-core and multi-processor computer systems may inherently be multi-threaded when integrated into multi-threaded operating systems such as Microsoft Windows XP Professional Edition (64-bit or 32-bit editions) and Linux. Can be.

In the embodiment shown in FIG. 12, each eight cores are responsible for only one tile, which generally operates independently of the other cores, each instance of x264 running on each other. PCI Express x1, based on a DVI capture card such as Sendero Video Imaging IP Development Board from Microtronix of Oosterhout, the Netherlands is used to capture uncompressed video at 640 × 480, 800 × 600, or 1280 × 720 resolution, Uses Direct Memory Access (DMA) to transfer captured video over the DVI bus to the system RAM. The tiles are arranged in a 4 × 2 arrangement 1205 (although in these embodiments they are 160 × 240 resolution, although they are shown as square tiles). Each instance of x264 is configured to compress one of eight 160 × 240 tiles, and in order to compress the I tiles followed by the seven P tiles shown in FIG. 12, they are each core after initial I tile compression. Enters one cycle, and each frame is synchronized after leaving the state with the others.

At each frame time, using the techniques described above, the resulting compressed tiles are combined into a packet stream and then the compressed tiles are sent to the destination client 415.

Although not shown in FIG. 12, if the combined 8 tile bit rate exceeds a certain peak rate 941, then all 8 x264 processes will have as many frames as needed until data for the combined 8 tile bit is transmitted. Pause for hours.

In one embodiment, client 415 runs like software on a PC running 8 instances of FFmpeg. The receiving process receives 8 tiles, each tile being sent to the FFmpeg instance, which stretches the tile and renders it for the appropriate tile position in the display device 422.

The client 415 receives input from the keyboard, mouse, and game controller from the input device driver of the PC and sends it to the server 402. The server 402 then applies the received input device data and applies it to a PC running Windows using an Intel 2.16 GHz Core Duo CPU, a game or application running on the server 402. The server 402 then generates a new frame and outputs it via the DVI output, and outputs it via the DVI output of the motherboard based graphics system, or the NVIDIA 8800GTX PCI card.

At the same time, the server 402 outputs the audio produced by the game or application via a digital audio output (e.g., S / PDIF), which is combined with the digital audio input of a dual quad core Xeon-based PC that performs video compression. do. Vorbis open source audio compressors compress simultaneously with video using any core that can use process threads. In one embodiment, the core performs audio compression first to finish compressing its tiles. The compressed audio is then sent along with the compressed video and expanded at the client 415 using the Vorbis audio extender.

Hosting Services Server Center Placement ( Hosting Service Server Center Distribution )

Light through the glass, such as an optical fiber, travels close to the speed of light in a vacuum so that the exact propagation speed for light in the optical fiber can be determined. In practice, however, allowing time for routing delays, transmission inefficiencies, and others, we observe that the optimal latency on the Internet reflects transmission rates close to 50% of the speed of light. Thus, the optimal 1000 mile round-trip latency is approximately 22 ms, and the optimal 3000 mile round-trip latency is approximately 64 ms, so a single server on the US coast can be too far away from the other shore (which can be much less than 3000 miles). The client will not be able to provide the required latency. However, as shown in FIG. 13A, if the hosting service 210 server center 1300 is located in the center of the United States (eg, Kansas, Nebraska, etc.), the distance to any location on the continental United States may be 1500 miles or less. And round trip internet latency is less than 32ms. Referring to FIG. 4B, although the worst-case latency allowed for ISP 453 users is 25 ms, typically we observe a latency close to 10-15 ms with DSL and cable modem systems. 4B also assumes a maximum distance of 1000 miles from the user premises 211 to the hosting center 210. Thus, a typical ISP user has a round trip latency of 15 ms and a maximum internet distance of 1500 miles is a round trip latency of 32 ms, such that the overall round trip latency is at a point where the user operates the input device 421 and sees a response on the display device 422. 1 + 1 + 15 + 32 + 1 + 16 + 6 + 8 = 80 ms. So an 80ms response time can typically be achieved over an internet distance of 1500 miles. This allows any user premises on the continental United States to be accessed with a short enough user ISP latency to a single server center centrally located.

In another embodiment shown in FIG. 13B, the hosting service 210 server centers HS1-HS6 are some large hosting service server centers located in close proximity to a very popular center and are strategically located throughout the United States (or Other geopolitical areas). In one embodiment, server centers HS1-HS6 exchange information over the Internet or a private network or a combined network 1301 thereof. With multiple server centers, the service provides low-latency for users with high user ISP latency 453.

Although distance on the Internet is certainly a contributing factor to round-trip latency over the Internet, sometimes other factors usually start activities that are not related to latency. Sometimes the packet stream is routed over the Internet to a distant location and back again, resulting in a long loop latency. Sometimes there is routing equipment on the path that is not working properly, which causes delays in transmission. Sometimes there is an overloaded traffic path, which leads to delays. Also, sometimes a failure occurs by preventing your ISP from sending to a given destination. Thus, a typical Internet is usually fairly fair from one point to another and has a latency determined by the best route and generally distance (especially with long distance links that result in routing out of the user's local area). It provides a connection, which ensures that reliability and latency are never guaranteed and often cannot be achieved on the general Internet from the user's premises to a given destination.

In one embodiment, user client 415 is initially connected with hosting service 210 to play a video game and use an application, and the client is each of the hosting service server centers HS1-HS6 available at startup. (Eg, using the techniques discussed above). If the latency is low enough for a particular connection, that connection is used. In one embodiment, the client communicates with all or a subset, and the lowest latency connection among the hosting service server centers is selected. The client will select the service center with the lowest-latency connection and the service center will be identified as one with the lowest-latency connection and provide this information (eg, in the form of an Internet address) to the client.

If a particular hosting service server center is overloaded and / or can withstand the latency for the hosting service server center where the user's game or application is lightly loaded, the client 415 is directed to another hosting service server center. In such a situation, the game or application in which the user is running will be stopped at server 402 at the user's overloaded server center and the game or application status information is sent to server 402 at another hosting service server center. Will be. The game or application will resume. In one embodiment, the hosting service 210 may naturally reach a breakpoint (eg, between levels in the game or after the user initiates a “save” command in the application) for the game or application to make a transfer. I will wait until. In another embodiment, the hosting service 210 will wait until the user activity ceases for a specified period of time (eg, one minute) and then begin transmitting at that time.

As discussed above, in one embodiment, the hosting service 201 subscribes to the Internet bypass service 440 of FIG. 14 to provide its clients with guaranteed latency. As used herein, an Internet bypass service is a service that provides a private network path from one point of the Internet with a guaranteed feature (eg, latency, transmission rate, etc.) to another point. For example, if hosting service 210 receives a large amount of traffic from a user using AT & T's DSL service provided by San Francisco, rather than routing to AT & T's San Francisco-based central office, hosting service 210 is hosted. For service 210, a high-performance personal data connection can be leased from a service provider (perhaps AT & T itself or another provider) between the San Francisco-based central office and one or more server centers. Then, if a route through the Internet that is common to San Francisco users using AT & T DSL from all hosting service server centers (HS1-HS6) results in too high latency, a personal data connection can be used instead. Although a personal data connection is generally more expensive than a route over a typical Internet, the impact of the overall cost will be low and users will experience a more consistent service because the connection between the user and the hosting service 210 will remain for too long at a small rate. .

Server centers often have two levels of backup power in the event of a power outage. Typically, the first tier backs up power from the battery (or alternatively readily available energy source, such as a flywheel connected to a generator and continues to operate), which immediately powers up and stops server center operation when the main power is lost. Keep it. If the power outage is simple, the main power returns quickly (eg within 1 minute), then the battery is required to maintain server center operation. However, if a power outage occurs for a long time, it can typically be operated for as long as the fuel possessed by the generator (eg diesel-power) and delivered for the battery. Such generators are very expensive because the server center is capable of producing as much power as can normally be obtained from the main power.

In one embodiment, each of the hosting services HS1-HS5 shares user data with the other, so if one server center goes down, it may stop running games and applications, and each server 402 may have a different server center. Sends game or application state data to servers 402 of the server and will notify each user's client 415 to communicate with the new server 402. If this usage occurs occasionally, the user will be allowed to deliver the user to a hosting service server center that cannot provide optimal latency (e.g., the user simply has to withstand high latency for the duration of the outage). It will allow many wide range choices for the user. For example, differences in time zones across the United States would allow users to go to bed at 11:30 pm on the east coast, while users at the west coast would begin the peak of video game use at 8:30 pm . If at the same time the West Coast Hosting Services Server Center is in power outage, there will not be enough West Coast Servers 402 at the other Hosting Services Server Center to handle all users. In such a situation, some users may be sent to a hosting service server center on the east coast with a server 402 available, and the result that follows the user will only be high latency. If a user is sent from a server center that loses power, the server center can be started in an interrupted sequence of its servers and appliances, and all appliances are stopped before the battery (or other immediate power backup) is exhausted. In this way, generator costs for server centers can be avoided.

In one embodiment, during heavy loading times of hosting service 210 (due to user loading peaks or one or more server centers become inoperable), users may choose to use other servers based on latency requirements of the games or applications they use. Is sent to the center. So a user using a game or application that requires low latency will benefit from the low latency server connections available when there is a limited supply.

Hosting Service Features ( Hosting Service Features )

FIG. 15 illustrates an embodiment of components of a server center for hosting service 210 used to illustrate the following features. Like the hosting service 210 shown in FIG. 2A, the components of this server center are controlled and coordinated by the hosting service 201 control system 401, unless otherwise limited.

Inbound Internet traffic 1501 from user client 415 is directed to inbound routing 1502. Typically, inbound Internet traffic 1501 will enter the server center via a high speed fiber optical connection for the Internet, but adequate bandwidth, reliability and low latency network connectivity will be sufficient. Inbound routing 1502 is a system switch of a network (the network may be implemented via an Ethernet network, a fiber channel network or any other means of transport) and a suitable application / game (“app / game”) server. (1521-1525) is a routing server that routes each packet and supports a switch that takes the arriving packet. In one embodiment, certain packets delivered to a particular application / game server represent a subset of the data received from the client and / or within the data center by other components (eg, networking components such as gateways and routers). Will be converted / changed. In some cases, for example, if a game or application is running on multiple servers in parallel, packets will be routed to one or more of the servers 1521-1525 at once. RAID arrays 1511-1512 are connected to inbound routing network 1502, and application / game servers 1521-1525 can read or write RAID arrays 1511-1512. Moreover, RAID array 1515 (which will be implemented as a multi-RAID array) is also associated with inbound routing 1502 and data from RADI array 1515 can be read from application / game servers 1521-1525. Inbound routing 1502 includes a tree-structured switch having inbound Internet traffic 1501 at the root; A mesh structure that interconnects all the various devices; Or as interconnected with a series of subnets with concentrated traffic among the devices that separate and exchange from the traffic concentrated between other devices; It can be implemented in a wide range of prior art network architectures. One form of network configuration is a SAN, and although typically used for storage devices, it can also be used for general high speed data conversion between devices. In addition, application / game servers 1521-1525 will each have multiple network connections with inbound routing 1502. For example, the server 1521-1525 will have a subnet and network connection attached to the RAID array 1511-1512 and another network connection with a subnet attached to other devices.

As already described with respect to the server 402 in the embodiment shown in FIG. 4A, the application / game servers 1521-1525 may all be configured identically, some differently, all differently. In one embodiment, when using a hosting service, each user typically uses at least one application / game server 1521-1525. For simplicity of explanation, it is assumed that a given user uses an application / game server 1521, but multiple servers may be used by one user, and multiple users may share a single application / game server 1521-1525. Can be. As described above, the user's control input sent from client 415 is received in inbound internet traffic 1501 and routed to application / game server 1521 via inbound routing 1502. The application / game server 1521 is used as a control input of a user's game or application running on the server and computes the next frame of video and audio associated therewith. Application / game server 1521 then outputs uncompressed video / audio 1529 to shared video compression 1530. The application / game server may output uncompressed video by some means, including one or more Gigabit Ethernet connections, but in some embodiments the video is output via a DVI connection and audio and other compression and communication channel status information may be output. Output via a universal serial bus (USB) connection.

Shared video compression 1530 compresses uncompressed video and audio from application / game servers 1521-1525. Compression will probably be performed entirely as hardware or as hardware on which software operates. There will be a dedicated compressor for each application / game server 1521-1525, but if the compressor is fast enough, a given compressor can be used to compress the video / audio from one or more application / game servers 1521-1525. . For example, at 60 fps the video frame time is 16.67 ms. If the compressor is able to compress one frame within 1 ms, the compressor gets input from each other at one server and then from another as many as 16 application / game servers 1521-1525. It can be used to compress the video / audio with a compressor that stores the state of the audio compression process and switches the context as cycled between the video / audio stream from the server. This substantially results in cost savings in compression hardware. In one embodiment, because different servers complete the frame at different times, the compressor resources are shared pool 1530 with shared storage means (eg, RAM, flash) to store the state of each compressor process. When the server 1521-1525 frames are complete and ready to be compressed, the control means determines if a compression resource is available at that time and compresses the frame of the uncompressed video / audio and the server to compress it. Provides a compressed resource with a state of processing.

Part of the state of the compression process of each server may be used as a reference for the P tile, and may include: advance frame extended frame buffer data, resolution of video output; Quality of compression; Tiling structures; Allocation of bits per tile; Note that it contains information about the compression itself, such as compression quality, audio format (eg, stereo, surround sound, Dolby ^® , AC-3). However, the state of the compression process also includes communication channel state information about peak rate 941, and whether a prior frame (as shown in FIG. 9B) is currently being output (and as a result the current frame may be ignored). In addition, whether or not channel characteristics exist in consideration of compression, such as potentially excessive packet loss, influences the decision for compression (eg, with respect to the frequency of the I tile). Since the peak bit rate 941 or other channel characteristics change over time, it is determined by the application / game server 1521-1525 that supports each user monitoring data sent from the client 415, so that the application / game server ( 1521-1525 send related information to shared hardware compression 1530.

In addition, shared hardware compression 1530 utilizes means such as those described above, applies FEC codes where appropriate, copies certain data, or extends to client 415 and extends to feasible stability and high quality. The compressed video / audio is packetized through other steps to adequately ensure the performance of the video / audio data stream received by it.

Some applications, such as described below, require the video / audio output of a given application / game server 1521-1525 that can be used simultaneously in multiple resolutions (or other multi-formats). If application / game servers 1521-1525 may be compressed in different formats, different resolutions, and / or different packet / error correction structures at the same time. In some cases, some compression resources may be shared among multiple compression processes that compress the same video / audio (eg, in many compression algorithms, there is a step of scaling to multiple sizes before the image is compressed by something). If different size images are required to be output, this step can be used to provide several compression processes at one time). In other cases, separate compression resources will be required for each format. In some cases, all the various resolutions and formats of the compressed video / audio 1539 requested for a given application / game server 1521-1525 will be output to outbound routing 1540 at a time. In one embodiment the output of compressed video / audio 1539 is in UDP format, and so is a unidirectional stream of packets.

Outbound routing network 1540 directs each compressed video / audio stream to an intended user or other destination through an outbound Internet traffic 1599 interface, which will typically be connected to the Internet and a fiber interface. And / or a set of routing servers and switches that return to delay buffer 1515, and / or back to inbound routing 1502, and / or output for video distribution through a private network (not shown). Outbound routing 1540 will output a given video / audio stream to multiple destinations at once. In one embodiment, this is done using Internet Protocol (IP) multitasking and a given UDP stream intended to be streamed to multiple destinations at one time is broadcast, and the broadcast is routed to the routing server at outbound routing 1540 and It is repeated by a switch. The plurality of destinations of the broadcast will be to the client 415 of the plurality of users over the Internet, and the plurality of application / game servers 1521-1525 via inbound routing 1502 and / or one or more delay buffers 1515. Will be. Thus, the output of a given server 1521-1522 is compressed into a single or multiple formats, and each compressed stream is directed to a single or multiple destinations.

Furthermore, in another embodiment, if multiple application / game servers 1521-1525 are used simultaneously by one user (eg, in a parallel processing configuration for generating 3D output of a complex scene), each server may have a result. Will generate a portion of the image, and the video output of multiple servers 1521-1525 can be combined by hardware compression 1530 shared into a combined frame, as if from a single application / game server 1521-1525. As it is, it is treated as described above at the forward point.

In one implementation, copies of all video generated by application / game servers 1521-1525 are written to delay buffer 1515 for at least a few minutes (15 minutes in one embodiment). This allows video "rewind" from each session to review each user's prior actions or usage (in the case of a game). Thus, in one embodiment, each compressed video / audio output 1539 stream routed to user client 415 is also multicast to delay buffer 1515. When video / audio is stored in delay buffer 1515, the directory of delay buffer 1515 locates the delay buffer 1515 where the delayed video / audio can be found and the source of the delayed video / audio (the application / game server). 1521-1525 for cross-reference.

Live, instant-visible, instant- To play  Games that can Live , Instantly -Viewable, Instantly-playable Games )

The application / game servers 1521-1525 are not only used to run a given application or video game for a user, but they also provide a hosting service 210 that supports navigation through the hosting service 210 and other features. It is used to create user interface applications. A screenshot of this user interface application, "game finder," is shown in FIG. This particular user interface screen allows the user to watch 15 games played (or delayed) live by other users. Each of the "thumbnail" video windows, such as 1600, is a live video window in motion that shows video from a user's game. The field of view shown on the thumbnail is the same as the field of view you are viewing, but it may be a delayed field of view (e.g. if you are playing a combat game, you do not want to be known to other users until some time where you hid, You will choose to delay any view of gameplay for a minute). Also, the field of view will be the field of view of the camera, unlike the field of view of any user. Through menu selections (such illustrations are not shown), the user will simultaneously select a game for viewing based on various criteria. With a small sampling of typical choices, the user plays one type of all games (all played by other players), only the top-ranked player, the player at a given level in the game, or a lower-ranked player (e.g., if a player Will learn the basics), "buddies" (or competitors) players, games with the majority of viewers, and so on (as shown in Figure 16).

In general, each user will decide whether others will be able to watch his or her game or application, and if others will be able to watch, he will decide whether to watch only the delay.

The application / game servers 1521-1525 that produce the user interface screen shown in FIG. 16 send messages to the application / game servers 1521-1525 while requesting whose game each user is. Requires audio feeds. The message is sent over inbound routing 1502 or other network. The message will include the size and format of the requested video / audio and will identify the user watching the user interface screen. A given user will choose to choose a "privacy" mode and no other user is allowed to watch the video / audio of his game (both his view or other), or as described in the previous paragraph, The user will choose to allow viewing the video / audio from her game, but the watching video / audio is delayed. If the user application / game server 1521-1525 receives and accepts a request to allow the video / audio to watch, it will approve the requesting server, which also requests the requested format or screen size (format and screen). The size will inform the shared hardware compression 1530 of the need to create a compressed video stream in addition to the assumption that it has already been created, and also indicate the destination (eg, the requesting server) for the compressed video. . If the requested video / audio is only delayed, the requesting application / game servers 1521-1525 will be notified, which is the location of the video / audio in the directory in the delay buffer 1515, the delayed video / audio's Delayed buffer 1515 will require delayed video / audio by finding the network address of the source application / game server 1521-1525. Once all these requests are generated and handled, up to 15 thumbnail-sized video streams are routed for application / game servers 1521-1525 to create user interface streams from outbound routing 1540 to inbound routing 1502. Will be expanded and displayed by the server. If the delayed video / audio stream is at too large screen size, application / game servers 1521-1525 will stretch the stream and reduce the video stream to thumbnail size. In one embodiment, the request for audio / video is a central “management” similar to the hosting service control system of FIG. 4A (not shown in FIG. 15) that directs the request back to the appropriate application / game servers 1521-1525. "" (And managed) to the service. Moreover, in one embodiment, no request is required because the thumbnails are "pushed" to the client of users who allow it.

At the same time all mixed audio from the 15 games will produce a dissonant sound. The user will select all sounds together in this way (probably getting a sense of "din" produced by all the actions shown), or the user will choose to listen to audio in one game at a time. Selection of a single game can be accomplished by moving the yellow selection box 1601 for a given game (yellow box movement can be accomplished by using the arrow keys on the keyboard, by moving the mouse, by moving the joystick, or by another device such as a mobile phone). By pressing a direction button). Once a single game is selected, only audio is obtained from that gameplay. Game information 1602 is also shown. For example, in the case of this game, the publisher logo (“EA”), the game logo, “Need for Speed Carbon” and the orange vertical bar indicate items related to the number of people watching or playing the game at a particular moment. Moreover, there are 145 players actively playing 80 different instances of Need for Speed Game (eg it can be played by individual player games or multiplayer games), and there are 680 observers (one of the users). "Stats" is provided to indicate. These statistics (and other statistics) are collected by the hosting service control system 401 and maintained in a RAID array to maintain a log of 210 hosting service operations, payments to moderately paid users, and publishers providing content. 1512). Some statistics are recorded due to operations by the service control system 401, and some are reported to the service control system 401 by the personal application / game servers 1521-1525. For example, application / game servers 1521-1525 running this game finder application may send a message to the hosting service control system 401 when a game is being watched (and when they start watching) to see how many games are watched. It will update you if it is. Some statistics may be used for user interface applications such as game finder applications.

If the user clicks the Activate button on their input device, they will see a thumbnail video zoomed up in the yellow box while remaining full screen. This effect is shown in the process in FIG. 17. Note that video window 1700 increases in size. To implement this effect, application / game servers 1521-1525 may be configured to execute an application / game server running a game selected to have a copy of the video stream for full screen size (at the resolution of the user's display device 422). 1521-1525) request the game routed for it. The application / game server 1521-1525 on which the game operates is shared that thumbnail-size copies of the game are no longer needed (unless other application / game servers 1521-1525 do not require such thumbnails). Notify hardware compressor 1530, which then sends a full screen size copy of the video to application / game servers 1521-1525 that enlarge the video. The user playing the game may or may not have the same resolution display device 422 as the user enlarges the game. Moreover, other observers of the game may or may not have a display device at the same resolution as the user magnifies the game (and will have other audio playback means, eg, stereo or surround sound). Thus, shared hardware compression 1530 may be used to determine whether a properly compressed video / audio stream has already been created to meet the needs of the user requesting the video / audio stream, and if present it is an application / game server that scales up the video. Inbound routing and application / game server 1521-1525 to zoom out video and compress other video copies suitable for the user, if not, to outbound routing 1540 to route a copy of the stream to 1521-1525. Notify outbound routing to send the stream back to 1502. The server that receives the full screen version of the currently selected video will stretch it and gradually scale it to full size.

18 shows what the screen looks like after the game has been enlarged to the full screen and the game is shown at the full resolution of the user's display device 422 as indicated by the image indicated by arrow 1800. The application / game server 1521-1525, where the game finder application is running, sends messages to other application / game servers 1521-1525 that provide thumbnails that are no longer needed and that other games are no longer being viewed. The message is also sent to the service control server 401. At that point, the display will only create an overlay 1801 on top of the screen that provides information and menu control to the user. As the game progresses, the audience grows to 2,503 spectators. With so many spectators, there are many spectators with display device 422 that have the same or nearly the same resolution (each application / game server 1521-1525 can scale the video for adjusting).

Since the depicted game is a multiplayer game, the user will decide when to join the game. The hosting service 210 will allow or disallow users participating in the game for a variety of reasons. For example, the user will have to pay to play the game and is not optional, and the user will not have enough ranking to participate in a particular game, or the user's internet connection will not have low-latency enough for the user to play. (E.g. there is no latency constraint for watching games, and games that are played far away (actually on other continents) can be viewed without worrying about latency, but to play the game, latency must be (a) to play the game) Be low enough for the user, and (b) are in the same position as other players with low-latency connections). If the user is allowed to play, the application / game servers 1521-1525 that provide the game finder user interface for the user may have the hosting service control server 401 load the game from the RAID array 1511-1512 to determine if the user is allowed to play. It will request to start (eg, browse and start) an appropriately configured application / game server 1521-1525 to play the game. And shared by the hosting service control server 401 to switch from compressing the video / audio from the application / game server hosting the game finder application to compressing the video / audio from the application / game server hosting the current game. Will command hardware compression 1530. The vertical sync of the game finder application / game service and the new application / game server hosting game is not synchronized, resulting in a time difference between the two sinks. Because the shared video compression hardware 1530 will begin compressing the video at the application / game servers 1521-1525 that complete the video frame, the first frame from the new server completes sooner than the full frame time of the original server. This will first be before the transmission of the compressed frame is complete (eg, consider the transmission time 992 of FIG. 9B: if uncompressed frame 3 963 is completed half the frame time earlier, it will be transmitted). Will affect time 992. In such a situation, shared video compression hardware 1530 will ignore the first frame from the new server (eg, as frame 4 964 is ignored 974), and the client ( 415) will maintain the last frame from the conventional server at the remaining frame time, shared video compression hardware (1530) will begin compressing the next frame time video from the new application / game server hosting the game, apparently to the user, the transfer from one application / game server to the other will look the same. 401 will then notify the application / game game server 1521-1525 hosting the game finder to switch to the idle state until it is needed again.

The user can then play the game. And, the exception is that the game will be immediately perceptually played (since it will be loaded from the RAID array 1511-1512 to the application / game game server 1521-1525 at gigabit per second), so the game is an ideal driver. It will be loaded into a server that is exactly the right for the game, with an operating system that has a registry configuration (in the case of Window) and that is configured correctly for the game without other applications running on the server that compete with the game's behavior.

Also, as the user progresses the game, each segment of the game will be loaded from the RAID arrays 1511-1512 to the server at a gigabit per second speed (e.g., 1 gigabyte loaded in 8 seconds), and the RAID array 1511. With the vast storage performance of 1512 (that's a shared resource among many users, it's very large and still cost effective), geometric setups or other game segment setups can be pre-computed and RAID arrays. 1512) and load very quickly. Moreover, because the hardware configuration and computer performance of each application / game server 1521-1525 are known, the pixel and vertex shaders can be precomputed.

Thus, the game will start almost immediately, it will work in the ideal environment, and the next segment will load almost immediately.

However, beyond this advantage, the user will be able to see others playing the game (via the game finder described above, or by any other means), and tips from watching if the game is interesting and if so Learns. And the user will be able to demo the game immediately without waiting for large downloads and / or installs, and the user will be able to play the game immediately, perhaps at a lower rate, or in the longer term Could be. And, the user will be able to play the game at home on a Windows PC, Macintosh, or television set, and can play the game while traveling or even on a mobile phone with a sufficiently low-latency wireless connection. And this can all be achieved without physically owning a copy of the game.

As mentioned earlier, the user may not allow anyone else to see his gameplay, may be able to see his game after a delay, or may be able to view his game by a selected user, You can make his game visible to all users. Nevertheless, in one embodiment for 15 minutes in delay buffer 1515, the video / audio will be stored, and the user will be able to "play back" and see his previous gameplay, pause, slow Play it back, play it fast forward, and so on, as he can while watching TV with a digital video recorder (DVR). Although in this example, the user plays the game, if the user uses the application, the same "DVR" performance is available. This can be helpful in reviewing preliminary work and in other applications as described below. Moreover, if a game is designed to have the ability to play back based on available game state information, the camera field of view may change, and then this "3D DVR" performance will also be supported, but it will not be possible for the game to support it. Will be required to be designed. The "DVR" performance using the delay buffer 1515 will work with any game or application of course, but for a limited amount of video generated when the game or application is used, but it is a user in the case of a game with 3D DVR performance. Can control "fly through" in the 3D of the "Preplayed Segment", and the resultant video is stored by the Delay Buffer 1515, and has the game state of the game segment record. Will be recorded as compressed video, but since the game status will also be recorded, other fly-throughs will be possible on later dates of the same segment of the game.

As described below, in hosting service 210 a user will have a respective user page, and they may inform themselves about themselves and other data. Among the things the user can tell is the video segment from the gameplay they are stored in. For example, if a user overcomes certain difficult challenges in the game, the user can "rewind" to just before the point of their great achievement in the game, and then make it visible to other users on the user's user page. Instructs hosting service 210 to store a video segment of some duration (eg, 30 seconds). To do this, the problem with application / game servers 1521-1525 is that the user will use a RAID array 1511-1512 to play the video stored in the delay buffer 1515 and then the video segment on the user's user page. To index.

If a game has the capabilities of a 3D DVR, as described above, then game state information requested for the 3D DVR can be recorded by the user and made available to the user's user page.

In an event where the game is designed to have a "spectator" (eg, the user can travel through the 3D world and observe motion without participating in it), add an active player, and the game finder application allows the user to It will allow players to participate in the game as well as the audience. In terms of implementation, there is no difference in the hosting system 210 even if the user is an audience instead of an active player. The game will be loaded into application / game servers 1521-1525 and the user will coordinate the game (eg, control the virtual camera to see the world). The user's gaming experience would be the only difference.

Multi-user collaboration ( Multiple User Collaboration )

Another feature of the hosting service 210 is that multiple users can collaborate while watching live video, even if they use very different devices to watch. This is useful when playing a game and when using an application.

Many PCs and mobile phones are equipped with video cameras and are capable of real-time video compression, especially when the image is small. Also, a small camera can be used to attach to a television, which is not difficult to compress in real time using software or many hardware compression devices to compress the video. In addition, many PCs and all mobile phones have microphones, and headsets are available as microphones.

These cameras and / or microphones, combined with local video / audio compression performance (especially employing the low-latency video compression techniques described herein), allow users to view video and / or from user premises 211 to hosting service 210. Audio may be sent along with input device control data. When such a technique is adopted, the performance values shown in FIG. 19 can then be achieved: the user can have his video and audio 1900 appearing on the screen in the game or application of another user. An example of this is a multiplayer game where teams compete in a car race. A user's video / audio can be selectively viewed and / or listened only to their team. And since there is no effective latency there, using the techniques described above may create conversations or actions between each other in real time without user perceiveable delay.

Video / audio integration may be accomplished by having compressed video and / or audio arriving at inbound internet traffic 1501 at the user's camera / microphone. Inbound routing 1502 sends video and / or audio to application / game game servers 1521-1525 that are allowed to view / listen to video and / or audio. The user of each application / game game server 1521-1525 selected to use video and / or audio compresses it and integrates it as desired by the game or application, as shown by 1900.

Although the example of FIG. 19 illustrates how such a collaboration is used in a game, such collaboration can be a tremendous power tool for an application. Given the situation in which large buildings are designed for New York City by architects of Chicago for real estate developers in New York, the decision is unfortunately to include financial investors in Miami Airport, satisfying investors and real estate developers. To do this, decisions need to be made about the building's design elements about how they match up with nearby buildings. Construction firms have high-resolution monitors with cameras attached to PCs in Chicago, real estate developers have laptops with cameras in New York, and investors have mobile phones with cameras in Miami. Construction firms can use hosting services 210 to host powerful architectural design applications with highly realistic 3D rendering capabilities, and can use many databases of buildings in New York City as well as the database of buildings under design. One architectural design application will run, but many of the application / game servers 1521-1525 require a lot of computer power. Each of the 3 users in a separate location will be connected to the hosting service 210, and each will be able to simultaneously view the video output of the architectural design application, but it will have the same device and network connectivity characteristics that each user has (e. G. 2560 × 1440 60fps display through an internet connection 20Mbps, a real estate developer in New York can see 1280 × 720 60fps images on his laptop via a 6 Mbps DSL connection, and the investor can 320 × via a 250Kbps cellular data connection to her mobile phone. 180 60fps images will be seen), which will be properly dimensioned by shared hardware compression 1520. Each party will hear the voice of the other party (the conference call will be handled by any of the many conference phone software packages available on the application / game server (s) 1521-1525), and the button operation of the user input device Thus, users will be able to use their local cameras to create videos that represent themselves. As the meeting progresses, architects appear to rotate the building with the same video and very photorealistic 3D rendering that every organization can see for the resolution of each organization's display device, making it to the side of the area where the other building is. You can make it look like it's been taken. It is not a problem that none of the local devices used by any organization have the ability to handle 3D animation with such realism, as well as storing or downloading the vast database required to render New York City's surrounding buildings. It doesn't matter.

At each user's point of view, despite their distance, even though they are separate local devices, they will simply have a smooth experience of incredible realism. And when an organization wants their face to communicate their emotional state better, they can convey it. Furthermore, if a real estate developer or investor wants to control a building program or use their own input device (which is a keyboard, mouse, keypad, or touch screen), they can control it and without perceived latency. Will respond (assuming their network connection does not have irrational latency). For example, in the case of a mobile phone, if the mobile phone is connected to a WiFi network at the airport, it will have very low latency. But if you use the cellular data network available today in the United States, it will probably have a noticeable rack. Still, the purpose of most gatherings is for investors to watch architects control building fly-bys or talk in videoconferences, although cellular latency can be acceptable.

Finally, at the end of the collaborative conference call, the real estate developer and investor will make their comment and terminate the hosting service, and the building company will “rewind” the conference video stored in the delay buffer 1515. rewind) and superficial representations and / or motions and opinions applied to the 3D model of the building created during the meeting. If there are special segments they want to store, the segments of video / audio can be moved from the delay buffer 1515 to the RAID arrays 1511-1512 for archive storage and later playback.

Also, from a cost point of view, if architects need to use computer-powered performance and a huge database in New York City for a 15-minute conference call, they have an expensive copy of a large database and own a very capable workstation. Rather, you only need to pay for the time you use the resource.

Video-Rich Community Services ( Video - rich Community Service )

Hosting service 210 enables an unprecedented opportunity to install video-rich community services of the Internet. 20 illustrates an exemplary user page for a game player in hosting service 210. Like the game finder application, the user page is an application that runs on one of the applications / game servers 1521-1525. All thumbnails and video windows on this page show continuously moving video (if the segments are short and they are repeated).

By using a video camera or uploading a video, a user (whose name is "KILLHAZARD") can post his own video 2000 that can be viewed by other users. The video is stored in a RAID array 1511-1512. Also, when another user comes to KILLHAZARD's user page, and if KILLHAZARD uses the hosting service 210 at that time, whatever he is doing is a live video 2001 (a user viewing his user page to watch him) Assuming he allows it). This depends on whether the KILLHAZARD is activated or, if so, whether it is the application / game server 1521-1525 he uses, the application / game server 1521-1525 hosting the user page application requested from the service control system 401. Will be achieved by Then, the same method is used by the game finder application, and the video stream and format compressed to the appropriate resolution is sent to the application / game server 1521-1525 on which the user pager application is running and it will be displayed. If the user selects the window of KILLHAZARD's live gameplay, then click on their input device as appropriate, and the window will expand (again, using the same method as the game finder application, depending on the characteristics of the user's Internet connection to watch). Appropriate and at the resolution of the viewing device's display device 422 the live video will fill the screen).

The major advantage of the prior art approach is that by viewing the user page, the user can watch live games not owned by the user, and can do very well without having a local computer or game console where the game is played. It provides a good opportunity for the user to see the game being played "in action" shown on the user page, which gives the viewer a chance to try or learn about the game they wish to do better. .

Video clips recorded or uploaded by cameras from friends of KILLHAZARD (2002) are also displayed on the user page, and the text under each video clip indicates whether the friend plays an online game (eg, six_shot is "Eragon" Game and MrSnuggles99 is offline, etc.). Clicking on a menu item (not shown) will switch the video clip of the friend from being recorded and uploaded to the live video of the friend playing the game in the hosting service 210 operating at their game moment. Thus, this will be a game finder for grouping friends. If a friend's game is selected and it is clicked, it will be expanded to full screen and the user will be able to watch the game being played full screen live.

Repeatedly, a user watching a friend's game does not have to own a local computing / game console resource to play a game or a copy of the game. Watching the game is effectively immediate.

As described above, when a user plays a game in the hosting service 210, the user can "rewind" the game and find the video segment he wants to save, and the video segment on his user page Can be stored. This is called "Brag Clips". The video segment 2003 is all Bragg clips 2003 stored by KILLHAZARD from the dictionary games he played. Number 2004 shows how many times a Bragg clip was viewed, and when the Bragg clip is shown, the user has an opportunity to rate them, and a number of orange keyhole shaped icons 2005 indicate how high the rating is. Show if it represents. The Bragg clip 2003 continuously loops according to the video remaining on the page when the user views the user page. When the user selects and clicks on one of the Bragg clips 2003, it becomes play, pause, rewind, fast-forward, stepped through, etc. Zoom in to show Bragg clip 2003 in accordance with the DVR controls allowing the clip.

Bragg clip 2003 playback is performed by the application / game server 1521-1525, which loads the compressed video segment stored in the RAID array 1511-1512 when the user records the Bragg clip, stretches it and plays it again. Is implemented.

In addition, Bragg clip 2003 may be a "3D DVR" video segment (eg, a game state sequence in a game that allows a user to change camera viewpoint and may be replayed) in a game that supports such functionality. The game state information in this case is stored in addition to the compressed video record of the particular "fly through" that the user makes when the game segment is recorded. When the user page is being shown, all thumbnails and video windows will loop continuously, and the 3D DVR Bragg clip 2003 will record Bragg clips recorded like compressed video when the user has recorded a "fly through" of game segments. (2003) will continue to loop. However, when the user selects a 3D DVR Bragg clip (2003) and clicks it, he adds a DVR control that allows the compressed video Bragg clip to be played, and the user clicks a button that gives them 3D DVR functionality for the game segment. You can do it. They will be able to control the camera "fly through" during their game segment, and if they want (and if the user with the user page allows it), they will be able to use the optional Bragg clip "fly through" in the form of compressed video. It will be able to record and then make it available to other observers of the user page (immediately, or after the owner of the user page has had the opportunity to review the Bragg clip).

This 3D DVR Bragg clip 2003 performance may enable the game to replay game state information recorded by other application / game servers 1521-1525. Since the game can be activated almost immediately (as described above), it is not difficult to activate the game limited to the game state recorded by the Bragg clip segment, and then to record the compressed video into the delay buffer 1515. Allow the user to "fly through" to the camera while Once the user has finished running, the "fly through" game is deactivated.

At the user's point of view, an active "fly through" with the 3D DVR Bragg clip 2003 eliminates the effort to control the DVR control of the linear Bragg clip 2003. They will not know how to play the game or anything about the game. They are just virtual camera operators who look at the 3D world during game segments recorded by others.

Users will also be able to multiplex their own audio with Bragg clips that have been recorded or uploaded from the microphone. In this way, Bragg clips can be used to create custom animations using characters and actions from the game. This animation technique is commonly known as "machinima".

As the user progresses through the game, they will achieve different levels of skill. The game being played reports the achievement to the service control system 401, and this skill level is posted on the user page.

Interactive  Animated ads ( Interactive Animated Advertisements )

Online advertising has changed from text to text, still to images, video, now interactive segments, and thin clients, typically Adobe Flash, have been running using animation. The reason that thin clients use animation is that users generally have some patience to be delayed with respect to the privileges of services or products sold to them. In addition, thin clients continue to operate with very low performance PCs, and advertisers have a high degree of confidence that interactive advertising can work properly. Unfortunately, animated thin clients such as Adobe Flash are limited in the amount of interactivity and duration of the experience (to ease download time).

FIG. 21 illustrates an interactive advertisement that allows the user to select the exterior and interior colors of the car while rotating around the car at the showroom, and real-time ray tracing shows how the car looks. After the user has selected an avatar to drive the car, the user can then take the car for a drive on a race track or to an exotic local like Monaco. The user can choose a larger engine, or a better tire, and can see how the modified configuration affects the performance of the car accelerating and sticking to the road surface.

Of course, the commercial is an effective sophisticated 3D video game. However, in order for these ads to be played on a PC or video game console, it will probably require 100MB of download, and in the case of a PC, it may require the installation of special drivers, and will not work if there is a lack of adequate CPU or GPU computing power. It may not. Thus, such advertisements are impractical in prior art configurations.

In the hosting service 210, these ads start almost instantly, run perfectly, and it does not matter what the user's client 415 performance is. So they are faster than thin client interactive ads, richer in experience, and very reliable.

Streaming geometry during the live animation ( Streaming Geometry During  Real-time Animation )

RAID arrays 1511-1512 and inbound routing 1502 provide RAID arrays 1511-1512 for reliable delivery of geometry immediately during application play or during gameplay during real-time animations (eg, fly-through to complex databases). And too low latency and too fast to design video games and applications that rely on inbound routing 1502.

With prior art systems, such as the video game system shown in FIG. 1, huge storage devices are available, especially in real home devices, which are too slow except in situations where the required geometry is somewhat foreseeable. Cannot be streamed. For example, a game driving on a special road can reasonably expect the geometry for a building to enter the field of view, and a huge storage device can find out where the upcoming geometry is located.

But in complex scenes with unpredictable changes (eg in battle scenes with complex characters around), if the PC or video game system has RAM completely filled with geometry for objects in the current field of view, the user suddenly has their characters Run their characters to see what's behind, and if geometry isn't preloaded into RAM, there will be a delay in displaying it.

In the hosting service 210, the RAID arrays 1511-1512 can stream data in excess of Gigabit Ethernet speeds and have an SNA network, which achieves 10 Gigabit per second rate over 10 Gigabit Ethernet or other network technologies. It is possible. Ten gigabits per second will load at least one gigabyte of data per second. At 60 fps frame time (16.67 ms), approximately 170 megabits (21 MB) of data may be loaded. Of course, even in a RAID configuration, rotating media will still experience greater latency than one frame time, but flash-based RAID storage will eventually be as large as a rotating media RAID array and this high latency will not occur. In one embodiment, large scale RAM write-through caching will provide very low-latency access.

Thus, mass storage, geometry with sufficiently high network speeds and sufficiently low latency can be streamed to application / game game servers 1521-1525 as fast as the CPU and / or GPU can process 3D data. . So, in the example given earlier, in order for the user to suddenly rotate their character and look backwards, the geometry for all the characters behind them may be loaded before the character's rotation is complete, so that for the user it may be as vivid as a live action. It would look like he or she is in the world of for realism.

As discussed earlier, one of the last limitations of photorealism computer animation is the human face, and since the sensitivity of the human eye is not perfect, a slight error from the photoreal face can cause a negative response from the observer. Fig. 22 Contour ^TM Reality Capture Technology (co-pending (co-pending) the subject of the application: "How to capture the behavior of the performers and the equipment" 10 / 942,609, September 2004, filed May 15; "capture the representation of the executors And a method and apparatus for improving marker identification in a motion capture system "11 / 066,954, filed February 25, 2005;" Shutter synchronization Method and apparatus for performing motion capture using ", 1 / 077,628, March 10, 2005;" Method and device for capturing motion using a random pattern on the capture surface "11 / 255,854, October 2005 Filed 20: "Method and System for Performing Motion Capture Using Phosphor Application Technology", 11 / 449,131, June 7, 2006; "System and Method for Performing Motion Capture Using Fluorescent Lamp Flash Effect" , 11 / 449,043, filed June 7, 2006; "Three Dimensions of Stop Motion Animation Characters Method and system for the application of ", 11 / 449,127, filed June 7, 2006, each of which is now assigned as the assignee of the current CIP application) how to bring the live performance to a very smooth capture surface and then It is illustrated with respect to the high polygon-count tracked on the surface (eg, the polygon motion follows exactly the motion of the face). Finally, the video of the live performance is mapped to the tracked surface to produce a textured surface and a photoreal result is generated.

Although current GPU technology can render polygons and surface textures and lights on multiple tracked surfaces in real time, and if the polygons and textures change every frame time (which will produce the most photoreal result), it is fast. It will consume the available RAM of modern PCs and video game consoles.

Using the streaming geometry technology described above, it will actually feed the geometry to the application / game game servers 1521-1525 continuously so that they can continuously animate the photoreal face and create a video game with the face. This makes it almost indistinguishable from a living face.

Interactive  Integration of linear content with Integration of Linear  Content with Interactive Features )

Movies, TV shows, and audio material ("linear content" can be widely used by home and office users in many forms). Linear content can be obtained from physical media such as CD, DVD, HD-DVD, and Blue-ray media. have. It can also be recorded as DVRs in satellite and cable TV broadcasts. And, it is available as pay-per-view (PPV) content through video on demand (VOD) on satellite and cable TV and cable TV.

Increasingly linear content is available over the Internet as download and streaming content Today there is no single place to experience all the features associated with linear media. For example, DVD and other video optical media generally have interactive features that are not available elsewhere, such as the commentary of a director who "makes" featurettes. Online music sites have cover art and song information that is not normally available on CDs, but not all CDs are available online. Websites related to television programming often hold comments from other features, blogs, and sometimes actors or staff.

Moreover, many video or sporting events have video games that are frequently released (in the case of videos), often with linear media (in the case of sports), or closely linked to real-world events (e.g. player trades, etc.) .

Hosting service 210 is well suited for the delivery of linear content that is linked together with individual forms of related content. Of course, the video provided will no longer require a highly interactive video game to deliver it, and the hosting service 210 can deliver linear content to a wide range of devices in a home or office or mobile device. FIG. 23 illustrates an exemplary user interface page for hosting service 210 showing the selection of linear content.

However, unlike most linear content delivery systems, the hosting service 210 also provides interactive components (eg Adobe Flash animation on a website (as described below), interactive overlays on HD-DVD, and DVDs on Menus and functions). Thus, the client device 415 no longer introduces a limitation with respect to being able to use its functionality.

In addition, the hosting system 210 can link dynamic video game content and linear content together in real time. For example, if a user is watching a Quidditch game in a Harry Potter movie, she can click a button and the movie will be stopped and immediately she will be moved to the Quidditch segment of the Harry Potter video game. After playing Quidditch, click the other button and the movie will resume immediately.

Photographically captured video with photoreal graphics and production technology is indistinguishable from live action characters, and as described here, when a user makes a move from a Quidditch game to a live action movie to a Quidditch game to a video game on a hosting service, In fact, the two scenes are indistinguishable. This provides new creative options that make it impossible for directors of linear content and interactive content (eg, video games) to distinguish between the two worlds.

Using the hosting service architecture shown in FIG. 14, control of the virtual camera in a 3D movie can be provided to the viewer. For example, in a scene that takes place within a train, it would be possible to look around the train during the story progression and allow the user to control the virtual camera. It can be used as ("assets") of all three-dimensional objects on the train by a reasonable level of computer power that can render scenes in real time as well as the original movie.

Even in non-computer generated entertainment, there is a very interesting interactive function that can be provided. For example, the 2005 film "Pride and Prejudice" has many scenes of gorgeous ancient English mansions. In certain mansion scenes, the user can pause the video and then control the camera for a tour of the mansion or surrounding area. To do this, a camera with a fish eye lens that tracks its location, Apple's prior art, could be transported through the mansion. QuickTime VR launches. The various frames are converted and the image is not distorted, stored in a RAID array 1511-1512 with the movie, and played back when the user makes a selection for a virtual tour.

Sports events, such as basketball games, and live sports events will be streamed through the hosting service 201 for users to watch, as they do on regular TV. After the user has watched a particular game, the video game of that game (the basketball player that eventually looks as photoreal as the actual player) starts with the player starting from the same location, and the user (perhaps each is controlled by one player) You can run it again to see if you can do better than the player.

The hosting service 210 described herein is well suited to support the future world because it allows it to withstand unrealistic computing power and mass storage resources when installed in a home or most office setting, and also in a home setting, While there are always older generations of PCs and video games, it keeps computing resources always up-to-date, the latest computing hardware is available, and all of this complexity of computing in hosting service 210 is hidden from the user, even though they From the user's point of view, even though it is using a very sophisticated system, from the user's point of view it is as simple as changing the channel on the television. Moreover, the user can access all computing power and the experience with computing power may come from any client 415.

Multiplayer game ( Multiplayer Games )

When the game is a multiplayer game, an application / game game server 1521-1525 and a network connected to the Internet and a game machine that does not operate in the hosting service 210 may be connected through an inbound routing 1502 network. When playing multiplayer games on a computer over a typical Internet, application / game game servers (1521-1525) have the advantage of very fast access to the Internet (compared to playing games on a server at home), but more It may be limited by the performance of other computers playing on a slow connection, and potentially also by the fact that it is a game server on the Internet designed to at least assign a common denominator, which will generally be the home computer of a slow consumer Internet connection.

However, if a multiplayer game is played entirely at the hosting service 210 server center, a different world can be achieved. Each application / game game server (1521-1525) that hosts games for users is the only server hosting central control for multiplayer games with very high speed, very low-latency and vast, very fast storage arrangements. But may be connected to other application / game game servers 1521-1525. For example, if Gigabit Ethernet is used as the inbound routing 1502 network, application / game game servers 1521-1525 communicate with each server and potentially play multiplayer games at gigabytes per second with a latency of 1 ms or less. It will communicate with any server hosting a central control. Moreover, RAID arrays 1511-1512 can respond very quickly and transfer data at gigabit per second. For example, if a user customizes a character with appearances and accoutrements, the character has unique behavior and a lot of geometry, and with prior art systems limited to game clients running home PCs or game consoles, When the character enters another user's field of view, the user waits for a long, slow download to complete and loads all geometry and behaviors data into their computer. Within the hosting service 210, the same download may go through the gigabit Ethernet provided in the RAID arrays 1511-1512 at a gigabit per second rate. If home users have an 8 Mbps Internet connection (which is very fast in today's standard), Gigabit Ethernet is 100 times faster. So a minute longer over a fast Internet connection would take less than a second over Gigabit Ethernet.

Top player groupings and tournaments ( Top Player Groupings and Tournaments )

Hosting service 210 is well suited for tournaments. Because if a game doesn't run on the local client, there's no chance for the user to cheat. In addition, because of the capability of output routing 1540 to multicast UDP streams, hosting service 210 can relay major tournaments to thousands of people at a time.

In fact, some video streams are so popular that when a large number of users try to receive the same stream (e.g., watching a major tournament), it is a content delivery network (CDN) such as Akamai or Limelight for mass distribution to many client devices 415. Will stream the video stream more efficiently.

A similar level of efficiency can be achieved when the CDN is used to display game finder pages that group top players.

In major tournaments, live celebrity announcers can be used to provide live broadcasts in any competition. Although a large number of users watch a major tournament, a relatively small number will play the tournament. Celebrity announcers' audio can be routed to application / game servers 1521-1525 that host users playing in the tournament and host observer mode copies of any game in the tournament, and audio can be recorded in multiples in addition to game audio. . The video of the famous announcer can be overlaid on the game, and possibly overlaid at the observer's point of view.

Accelerate web page loading ( Acceleration of Web Page Loading )

The major transport protocol of the World Wide Web, Hypertext Transfer Protocol (HTTP), was understood and defined in the era when only businesses had high-speed Internet connections and consumers who were online used dial-up modems or ISDN. At the same time, the "gold standard" of fast connection is a T1 line that transmits 1.5 Mbps data at the same time (eg with the same amount of data in both directions).

Today, the situation is completely different. In many advanced worlds, the average home connection speed via DSL or cable modem is considerably higher than the T1 line. In fact, in some parts of the world, fiber to the curb assumes data rates as fast as 50 to 100Mbps.

Unfortunately, HTTP has not been designed to make this dramatic speedup an effective advantage (it has never been realized). A website is a collection of files on a remote server. Simply, HTTP requests the first file, waits for the first file to download, then requests the second file and waits for the file to download. In fact, HTTP allows more than one "open connection", for example, allowing more than one file to be requested at a time, but because of the agreed-upon standard (and protecting the web server from overload) Only a few open connections are allowed. In addition, since the path of the web page is constructed, the browser often does not recognize a number of simultaneous pages that are available for download immediately (eg, after parsing the page's phrase, the image that needs to be downloaded). New files will appear. Thus, files on a website are loaded one at a time essentially because of the request-response protocol used in HTTP, with 100 ms latency associated with each file being loaded (access to a typical web server in the United States).

With a relatively low speed connection, it doesn't cause much trouble because the download time of the file itself takes precedence over waiting for the web page. However, as connection speeds increase with complex web pages, problems begin to arise.

In the example shown in FIG. 24, a typical commercial website is shown (this particular website is of a major sneaker brand). This website has 54 files. The files include HTML, CSS, JPEG, PHP, JavaScript and Flash files, and video content. The total 1.5 Mbytes must be loaded before the page is activated (ie the user can click on it and be available). There are many reasons for having a large amount of files. Firstly, it is a complex and sophisticated web page, and another reason is to dynamically gather information based on the information of the user accessing the page (i.e. in which country the user is, in what language, and the user has purchased the item before) And so on, depending on these factors, other files are downloaded. Still, this is a typical commercial web page.

Fig. 24 shows the total amount of time that elapses before the web page is opened as the connection speed increases. With a 1.5Mbps connection speed 2401, using a traditional server with a traditional web browser, it takes 13.5 seconds to open a web page. With a 12Mbps connection speed (2402), load time can be reduced to 6.5 seconds or about twice as fast. But with 96Mbps connection speed (2403), load time is reduced to only 5.5 seconds. The reason is that at such high download speeds, the time to download the file itself is insignificant, but the latency per file still remains approximately 100ms each, resulting in 54 files * 100ms = 5.4 seconds of latency. Therefore, no matter how fast your home connection is, it will always take at least 5.4 seconds for this website to be active. The other factor is server-side queuing; Every HTTP request is added after queuing, so on a busy server this will have a noticeable impact because of all the small items you get from the web server, and HTTP makes a need to wait for its sequence.

One way to solve this issue is to discard or redefine HTTP. Alternatively, a website owner may want to consolidate multiple files into a single file (eg, in Adobe Flash format). But the real problem is that not only the company but also many others invest a lot in building the website. Moreover, if some homes are connected at 12 to 100 Mbps, the majority of homes are connected at low speeds, and HTTP works well at low speeds.

One alternative is to host a web browser on the application / game servers 1521-1525, host the files of the web server on the RAID array 1511-1512 (or potentially an application that hosts the web browser in RAM or On local storage on game servers (1521-1525), with very fast interconnects via inbound routing (1502) (or local storage), using HTTP rather than showing 100 ms latency per file using HTTP. There will be a minimum latency per file. Then, instead of having the user in her home accessing the web page via HTTP, the user can access the web page through the client 415. Then, with a 1.5 Mbps connection (because this website does not require much bandwidth for its video), the web page may be active at less than 1 second per line (2400). In essence, there will be no latency until the web browser operating on the application / game servers 1521-1525 displays the page, and there will be no noticeable latency until the client 415 displays the video output from the web browser. As the user's mouse moves and / or types on the webpage, the user's input information will be sent to a web browser running on the application / game servers 1521-1525, and the web browser will respond accordingly.

The disadvantage of this approach is that if the compressor is constantly transmitting video data, bandwidth is used even if the webpage is fixed. This can be solved by configuring the compressor to send data only when the web page changes, and to send data to some of the changing pages if the data is sent. On the other hand, there are websites with several constantly changing flash banners, such websites tend to be upset, and usually web pages are stationary unless they cause something to move (video clips). For such a website, less data will be sent using the hosting service 210 than a traditional web server because only the images that are actually displayed are sent, there is no code that the thin client can execute, and a rollover image. There is no huge object like, never seen.

Thus, using the hosting service 210 to host a legacy web page, the web page consumption time can reduce where the web page is opened as if changing the channel on the television: the web page is effectively activated immediately. do.

Game and Of application  Possible debugging ( Facilitating Debugging of Games  and Applications )

As mentioned above, video games and applications with real-time graphics are very complex applications and typically they are brought to market with bugs. Even if the software developer gets feedback from the user about the bug, it means returning the machine after the crash, which makes it difficult to clearly determine what caused the game or real-time application to crash or behave improperly.

When a game or application runs on hosting service 210, video / audio output of the game or application is continuously written to delay buffer 1515. Moreover, the monitoring process operates each application / game server 1521-1525 to periodically report to the hosting service control system 401 whether the application / game servers 1521-1525 operate reliably. If the monitoring process fails to report, then the server control system 401 will attempt to communicate with the application / game servers 1521-1525 and if successful collect machine state not available. Any information available with the video / audio recorded by the delay buffer 1515 will be sent to the software developer.

Thus, the game or application software developer gets a notification of the failure from the hosting service 210 and gets a record for each frame that caused the failure. This information is very valuable for finding and fixing bugs.

In addition, when the application / game servers 1521-1525 fail, the server restarts at the most recent restart point and provides the user with an apologiaing message about technical difficulties.

resource  Share and reduce costs Resource Sharing and Cost Savings )

The system shown in FIGS. 4A and 4B provides various benefits to users and game and application developers. For example, home and office client systems (eg, PCs or game consoles) typically use only a small percentage of time per week. "Active Gamer Benchmark Study" released on October 5, 2006 by Nielsen Entertainment (http://prnewswire.com/cgi-bin/stories.pl?ACCT=104&STORY=/www According to / story / 10-05-2006 / 0004446115 & EDATE =), active gamers spend an average of 14 hours a week playing video game consoles and using handheld computers about 17 hours a week. The data also noted that for all game activities (including consoles, hand-helds, and PC games), active gamers spend an average of 13 hours a week. Taking a high percentage of the time playing console video games, 168 hours a week (24 hours * 7 days). This means that at the home of an active gamer, video game consoles use only 17/168 = 10% of the time of the week. Or 90% of the time, no video game console is used. Given the high cost of video game consoles, the fact that manufacturers subsidize such devices is inefficient use of expensive resources. In business, PCs are also used for only a small amount of time per week, and non-portable desktop PCs often require the best applications, such as Autodesk Maya. Although some businesses are available at all times and on holidays, some PCs (eg, portable PCs brought home for work in the evening) are used at all times and on holidays, and most business activities are on Monday to Friday. Desktop PC utilization tends to follow this uptime because it tends to be concentrated at 9:00 am to 5:00 pm, less use during holidays and breaks (such as lunch), and most PC usage occurs. If we assume that the PC is used continuously from 9 to 5 o'clock, five days a week, the PC is used for 40/168 = 24% of the time of the week. High-performance desktop PCs are a very expensive investment for office, resulting in very low levels of utilization. Schools that teach using the desktop use less time a week, and most teaching is done during the day, from Monday to Friday, even if the teaching time varies. Thus, PCs and video game consoles are typically used for only a small amount of time per week.

Note that many people use computers in the office or at school during the day of the week, Monday through Friday, not on holidays, and they usually do not play video games during this time. And when playing video games, it will usually be at different times: evenings, weekends, and holidays.

Looking at the type of hosting service shown in FIG. 4A, the utilization pattern is the result of the very efficient use of resources, as described in the two paragraphs above. In particular, if a user requires a real time response of a complex application such as a sophisticated 3D video game, the number of users that can be provided by the hosting service 210 at a given time will be limited. However, unlike a video console at home or a PC used for business that is not used most of the time, server 402 can be reused by other users at different times. For example, a high-performance server 402 with high-performance dual CPUs, dual GPUs, and large amounts of RAM can be used in offices and schools between 9 am and 5 pm on weekdays, but can play sophisticated video games on evenings and weekend holidays. It is utilized by gamers. Similarly, low performance applications may be utilized during working hours by Celeron CPUs, low or no GPUs, low performance GPUs, limited RAM, and low performance servers during the off-hours period. Can be used.

Moreover, with the deployment of hosting services described here, resources are efficiently shared among thousands, if not millions, of users. In general, online services have only a small percentage of the total users using the service at any given time. If you consider the Nielsen video game usage statistics earlier, it's easy to see why. If an active gamer plays console games for 17 hours a week, we have evenings (5am to 12am, 7 hours * 5 days = 35 hours / week) and weekends when the peak hours of the game are off hours. Assuming (8 am-12 am, 16 hours * 2 = 32 hours / week), there is a peak time of 65 (35 + 32) hours at 17 hours of game time. The exact peak user loading of the system is difficult to approximate for many reasons: some users will play during non-peak hours, and when the users form groups, it will most likely be day time, and the peak time zone will affect the type of game (e.g., Children tend to play games early in the evening). However, the average time gamers play the game is much less than the daytime gamers play the game, and at a given time the hosting service 210 uses only a small number of users. For the purposes of this analysis, we will assume that the peak load is 12.5%. Thus, only 12.5% of computing, compression, and bandwidth resources are used at any given time, and only 12.5% of hardware costs result in supporting a given user to play a given level of performance game due to resource reuse.

In addition, if some games and applications require more computing power than others, resources will be dynamically deployed based on the game you are playing and the application you are running. Thus, a user who selects a low performance game or application will be assigned to a low performance (less expensive) server 402, and a user who selects a high performance game or application will be assigned to a high performance server 402 (more expensive). Indeed, a given game or application has a low performance and a high performance portion of the game or application, and the user can choose between the portions of the game or application to keep the user running on the lowest price server 402 as needed for the game or application. It is possible to exchange one server 402 with another server 402. A much faster RAID array 405 than a single disk is also available in the low performance server 402, which has the advantage of speeding up disk transfers. Thus, depending on the game you are playing or the application you are using, the average cost per server 402 is much lower than the price of the most expensive server 402 of a high performance game or application, but even the low performance server 402 may be a RAID. Disk performance benefits will be obtained from the arrangement 405.

Furthermore, the server 402 of the hosting service 210 is nothing more than a diskless PC motherboard or a peripheral interface other than the network interface, but will only be integrated down to the SAN 403 and a single chip with a fast network interface. In addition, RAID array 405 will be shared by more users than disks, and active per user disk cost will be less expensive than a single disk drive. All these devices will be racked in an environmental control server room environment. If server 402 fails, it can be easily repaired or replaced at hosting service 210. In contrast, an independent device that must withstand a reasonable mechanical loss from the stiffness, hitting, or dropping of a PC or game console in the home or office must have a housing (cover), have at least one disk drive, and adverse environmental conditions. It is sold by a retailer who receives a distribution margin because it must withstand (for example, if it is to be put in an overheated AV cabinet with another device) and must be packaged and shipped. In addition, even if a low-performance game or device is operated most of the time (or a section of a game or application), the PC or game console will be able to achieve the best performance of the computer or game that was intensely expected at some point in the future. Environment settings should be made. And if a PC or console fails, repairing it is an expensive and time-consuming process (which adversely affects manufacturers, users and software developers).

Thus, the system shown in FIG. 4A provides an opportunity for a user to compare with the resources of the local computer at home, office, or school to experience a given level of computer performance, and provides computing performance through the architecture shown in FIG. 4A. Would be much cheaper.

Eliminate the need for upgrades ( Eliminating The Need to Upgrade )

Moreover, users no longer have to worry about upgrading their PCs or consoles to handle new games or new high-performance applications. Regardless of what type of game or application the server 402 requires, the hosting service 210 allows the user to use any game or application almost immediately (i.e., RAID array 405 or server 402). Loads local storage quickly, properly updated and corrects bugs (ie, software developers can choose to configure server 402 to an ideal environment for a given game or application, with optimal drivers) With the configuration of the server 402 and over time, the developer immediately provides updates and fixes bugs for all copies of the game or application of the hosting service 210). In fact, after the user has started using the hosting service 210, the user knows that he has provided and started a better experience for games and applications (e.g., through updates or bug fixes), which probably did not exist before a year ago. It may be the case that a user discovers a new game or application made available for a service 210 utilizing a computing technology (eg, a high performance GPU) after one year. Therefore, it is impossible for a user to purchase the technology after one year of playing a game or operating an application. Because the computing resources that play games and operate devices are invisible to the user (i.e., at the perception of the user, the user simply selects a game or application that works almost immediately as much as changing a television channel). It will be "upgraded" without recognizing it.

Eliminate the need for backup Eliminating the Need for Backups )

Another major problem users face in business, school and home is backup. Information stored on a local PC or video game (e.g., user's game achievement and ranking in the case of a console) may be lost due to error or unintended deletion of the disk. Manual or automatic backups are provided for PCs that can be used for many applications, and game consoles can be uploaded to an online server for backup, but local backups are typically copied to another local disk that is stored somewhere secure and organized ( Or other storage devices that do not lose data), backups via online services are often limited because of the slow upstream speeds typically through low-cost Internet connections. Data stored in the RAID array 405 with the hosting service 210 of FIG. 4A can be configured using well known RAID configuration prior art, which will not lose any data if the server center fails. The technician will notify you of the failed disk, replace it with the disk that automatically performed the update, and the RAID array will tolerate the failure once more. In addition, disks backed up on a regular basis to a secondary storage location where all of the disk drivers are close together, with a fast local network between them via the SAN 403, and can be relocated and stored in a server center or away. It's not difficult to deploy everything in the server center. From the point of view of the user viewing the hosting service 210, the data is always simply protected and never has to think about backup.

Access demo Access to Demos )

Users often want to try before buying a game or device. As mentioned above, there are prior art techniques for demonstrating games and applications (demo language means trying a demo version, also called a demo (not a noun)). Each of them suffers from limitations and / or inconveniences. It is easy and convenient for users to try the demo when using the hosting service 210. In fact, every user selects and tries a demo through the user interface (as described below). The demo will load almost immediately on the appropriate server 402 and will run like any other game or application. Whether the demo requires a high performance server 402 or a low performance server 402, whether using the home or office client 415 from the user's point of view, the demo will work. Game or application demo software makers can control exactly which demos are allowed for how long and, of course, demos contain user interface elements that give users the opportunity to gain access to the full version of a game or application. Should be

Because demos are often less than cost or free, some users will use the demos repeatedly (especially game demos that play repeatedly). The hosting service 210 can apply various techniques to restrict a given user from using the demo. The simplest approach is to set a user ID for each user and limit the number of times a given user ID is allowed to run the demo.

But the user will set multiple user IDs, even if it is free. One technique to address this problem is to limit the number of times a given client 415 is allowed to run a demo. If the client is a standalone device the device will have a serial number and the hosting service 210 can limit the number of times a demo can be handled by a client of that serial number. If the client 415 is operated as software on a PC or other device, the serial number may be assigned by the hosting service 210 and stored on the PC and used to limit the demonstration usage, but the user may reprogram the PC. The serial number may be erased or changed, and other countermeasures are provided by the hosting service 210 to record the PC network expansion card, MAC address (and / or other mechanical special identifiers such as the hard drive serial number, etc.) and use the demo. To limit. However, assuming that the MAC address of the network expansion card changes, this is not a safe way to fail. Another approach to limiting the number of uses of the demo is to allow it to run with a given IP address. Although the IP address is periodically reassigned by the cable modem and DSL provider, this does not happen very often, and a small number of demonstration uses are typically assumed given the IP in the area for residential DSL or cable modem access is determined. Can be set. There may also be multiple mechanical devices in a home under a NAT router sharing the same IP address, but in a typical residential environment there will be a limited number of such devices. If the IP address is at one company, the majority of the demo can be set up for one company. But finally, the combination of all the previously mentioned approaches is the best way to limit the number of demonstrations on your PC. Although there is no firm, technical, and unsuccessful way to limit an experienced user from repeating the use of the demo, there are enough obstacles to create so many restraints that the majority of PC users don't endeavor to abuse the demo system. It can be created, and users use demos rather than tending to try new games and applications.

school, business  And benefits from other organizations ( Benefits to Schools , Businesses and Other Institutions )

Significant advantages accumulate in businesses, schools, and other organizations that utilize the system in which FIG. 4A is shown. Offices and schools are expensive to install, maintain and upgrade, especially when it comes to PCs running high-performance applications like Maya. As mentioned earlier, PCs are generally only used a few hours per week, and the price of a PC for a given level of performance is much higher in an office or school environment than in a server center environment.

For large businesses or schools (eg, large universities), this can be practical in IT departments such as entities that build server centers and maintain computers with remote access over LAN-level connections. There are a number of solutions for accessing LAN or personal high bandwidth connections to remote computers between offices. For example, through a virtual network computing device such as Microsoft Windows Terminal Server or RealVNC's VNC, or Sun Microsystems' thin client, users can gain remote access to a PC or server with a range of qualities of graphical response time or user experience. It becomes possible. In addition, such self-managed server centers are typically dedicated to one such place, such as a business or school, and take advantage of the possible redundancy when completely different applications (such as entertainment and office equipment) use the same computer resources at different times of the week. Can't. Thus, many businesses and schools lack the scale, resources, or skills to build server centers that have LAN-speed network connections for each user. In fact, a large percentage of schools and businesses have the same Internet connections as homes (eg DSL, cable modems).

However, such organizations may still need a very high performance computing on a regular or periodic basis. For example, a small architectural firm will have relatively few architects and will usually have modern computing needs when designing, but will periodically require very high-performance 3D computing (eg new architectural design for customers). When creating a three-dimensional fly-through). This system is well suited for such an organization as in FIG. 4A. The organization does not have any needs beyond the same kind of network connection (eg DSL, cable modem) provided in the home, and is typically very inexpensive. They may utilize an inexpensive PC, such as client 415, or may utilize a dedicated, inexpensive device that provides the PC all at once and executes control signal logic 413 and low-latency video stretching 412. This feature is attractive for schools that have sensitive component damage or PC theft problems within the PC.

This arrangement solves many problems for such an organization (and many of these advantages can also be shared by home users with general purpose computing). Operating costs (which ultimately have to go through some form to the user to do a viable business) can be very low, because (a) computing resources are shared with applications that have different peak hours of use during the rest of the week, (b) the organization can gain access (and incur cost) to high-performance computing resources only when needed, and (c) the organization does not have to provide resources for backup to maintain high-performance computing resources.

Removal of piracy ( Elimination of Piracy )

In addition, games, applications, interactive movies, and the like can no longer be pirated as they are today. Because the game runs in the service center and users are not provided access to the underlying program code, there is no piracy there. Even if you duplicate the source code, you won't be able to run it on a standard game console or home computer. This opens up markets in places like China where standard video gaming is not available. Resale of your game is not possible.

For game developers, there is little market discontinuity as it is today. The hosting service 210 can be updated over time as the gaming needs change over time, in contrast to the current situation where an entirely new generation of technology is upgraded by users and developers and game developers rely on timely delivery of hardware platforms. There will be.

streaming Interactive  video( Streaming Interactive Video )

The above description provides a broad range of applications that can be achieved by a new fundamental concept of general Internet-based, low-latency streaming interactive video, which implicitly includes audio with video, as used herein. Prior art systems that provide streaming video over the Internet can be activated with applications that can only run with high latency interactions. For example, basic playback controls for linear video (eg, stop, rewind, fast forward) operate with moderately high latency, which makes it possible to choose between linear video feeds. And, as noted above, some video game features allow them to be played with high latency. However, the high latency (or low compression ratio) of prior art approaches for streaming video severely limits the potential application of streaming video or narrows their deployment for specialized network environments, and even in such environments, prior art Practically burdening the network. The technology described herein opens the door for a wide range of applications that can have low-latency streaming interactive video, which may be possible over the Internet, in particular over consumer-grade Internet connections.

In fact, a client device as small as the client 465 of FIG. 4C is sufficient to provide an enhanced user experience with very fast networking and any amount of fast storage, any amount of computing power between powerful servers. May allow computing of the criteria. Moreover, because bandwidth requirements have not grown as much as the computing power of system growth (e.g., bandwidth requirements are tied only to display resolution, quality, and frame rate), broadband Internet connectivity is ubiquitous (e.g., widespread low-latency). Wireless coverage, reliable, high enough bandwidth to meet the needs of all user's display devices 422, and a thick client (PC running Windows, Linux, OSK, etc.) The question of whether a mobile phone) or a thin client (such as Adobe Flash or Java) is necessary for typical consumer and business applications.

The advent of streaming interactive video is rethinking assumptions about the architecture of computing architectures. This example is the hosting service 210 server center embodiment shown in FIG. 15. The video path for delay buffer and / or group video 1550 is selectable via multicast streaming interactive video of application / game servers 1521-1525 via path 1552 in real time or via path 1551. The feedback loop returns to the application / game servers 1521-1525 after a delay. This may allow for a wide range of practical applications that are impossible or impossible to execute via prior art servers or local computing architectures (eg, as shown in FIGS. 16, 17 and 20). However, as a more general architectural feature, what the feedback loop 1550 provides is a streaming interactive video-level circulation because the video can loop back indefinitely as the application requires. This may make the possibility of a wide range of applications never available before.

Another major architectural feature is that the video stream is a one-way UDP stream. This can effectively enable any degree of multicasting streaming interactive video (in contrast, bidirectional streams, i.e. bidirectional streams such as TCP / IP streams, are increasingly getting back-and-forward to the network as the number of users increases). You will be able to create more traffic logjams). Multicasting is an important performance in server centers because it responds to the growing need of Internet users (and indeed the world's population) to communicate one-to-many or many-to-many. Because it allows the system to In other words, as in FIG. 16, the streaming interactive video recursion and multicasting examples discussed here are the tip of a large iceberg.

In one embodiment, the various functional modules illustrated herein and related steps include hardware logic by an on-demand semiconductor (ASIC) or a combination of programmed computer components and user-defined hardware components to perform the steps. It can be done by special hardware components.

In one embodiment, the module may be implemented with a programmable digital signal processor ("DSP"), such as Texas Instruments' TMS320x architecture (eg, TMS320C6000, TMS320C5000, ..., etc.). Various DSPs may still be used while following this fundamental theory.

The embodiment may include various steps as described above. The steps may be implemented by machine-executable instructions such that a general purpose or special purpose processor may perform any step. Various components not related to this basic theory, such as computer memory, hard drives, and input devices related to these basic principles, are left out of the drawings in order to avoid obscuring the appropriate aspects.

The problem of the disclosed subject matter may also be presented as a machine-readable medium for storing machine-executable instructions. Machine-readable media may be flash memory, optical disks, CD-ROMs, DVD-ROM RAMs, EPROMs, EEPROMs, magnetic or optical cards, propagation media, and other suitable for storing electronic instructions. Forms of machine-readable media, and the like. For example, the present invention relates to a computer that makes a request from a remote computer (e.g., a server) by way of a data signal contained in another propagation medium or carrier wave via a communication link (e.g., modem, network connection) ( For example, a computer program that can be transferred to a client).

In addition, this may be provided as a computer program product comprising a mechanically readable medium having stored instructions that may be used to perform a computer program to perform an operator of a sequence in which the elements in question of the disclosed subject matter (eg, a processor or its Out electronic devices). Alternatively, the operation may be performed in a combination of hardware and software. Machine-readable media include machine-readable media suitable for storing floppy diskettes, optical disks, magneto-optical disks, ROM, RAM, EPROM, EEPROM, magnetic or optical cards, propagation media or other forms of media / electrical instructions. May include, but is not limited to, media. For example, a problem element of the disclosed subject matter may be downloaded like a computer program product, which program may be a remote computer for requesting processing in accordance with the manner of a data signal contained in other propagation medium or carrier wave via a communication link or May be sent to the electronic device (eg, a modem or network connection).

Additionally, although the problems of the disclosed subject matter have been described in connection with specific embodiments, many variations and substitutions are well within the scope of this disclosure. Accordingly, the specification and drawings are to be regarded in an illustrative sense rather than a restrictive sense.

205: client device, 206: internet,
210: hosting service, 211: user premises,
220: software developer, 401: hosting service control system,
402: server, 403: storage area network,
404: video compression logic, 413: control signal logic,
415: client, 422: display device,
462: Ethernet jack, 479: Bluetooth input device.

Claims (3)

Dividing the program code and / or data used to execute the online application into a first type and a second type;
Storing the first type of program code and data in a first type of memory that provides relatively low latency memory access;
Storing the second type of program code and data in a second type of memory that provides relatively high latency memory access compared to the first type of memory;
Retrieving program code and data from the first and second types of memory in response to a client request to execute an online application; And
And sending a streaming interactive video stream representing the image generated by the application to a client. The method according to claim 1,
And said first type of memory is a flash memory and said second type of memory is a hard drive. The method according to claim 1,
And said application is a video game.

Images