KR19980017782A - Stop-and-Go Queue Service Using EDD in ATM Networks - Google Patents

Abstract

In the stop-and-go queue service method in which a packet is mounted in a frame and delivered through a path in a network to guarantee end-to-end delay and delay jitter, the composite frame size of the output link is configured to be an integer multiple of the reference frame. In the stop-and-go queue service method using EDD in ATM network, which has better bandwidth density and increased bandwidth usage efficiency by improving various delay limits and bandwidth allocation characteristics. Relate to the size of a first type frame at a time interval T1, a time interval T2 of a second type frame to k2T1, a time interval T3 of a third type frame to k3T1, and a fourth After setting the time interval T4 of the type frame to k4T1 and the time interval Tn of the nth type frame to knT1 (where 1 <k2 <k3 <k4 <… n), each frame is terminated. By applying the EDD at the time it is characterized by being a method for selecting a cell to be served.

Description

Stop-and-Go Queue Service Using EDD in ATM Networks

The present invention relates to a stop-and-go queue service method using EDD in an ATM network. In particular, the present invention relates to a stop-and-go that delivers packets through frames in a network by mounting packets in frames to ensure end-to-end delay and delay jitter. In the Go Queue service method, the composite frame size of the output link is composed of integer multiples of the reference frame, so that various delay limits and bandwidth allocation characteristics are improved, resulting in better bandwidth density and increased bandwidth utilization to accommodate more calls. The present invention relates to a stop-and-go queue service method using EDD in an ATM network.

Meanwhile, the international telecommunication union-telecommunication standardization sector (ITU-TS) adopts an asynchronous transfer mode (ATM) as a technology to effectively promote a broadband integrated information communication network, and the ATM network uses all information. Since the packet is a packet of limited sized information units called cells, and is transmitted by statistical multiplexing, it is possible to flexibly accommodate various different services and efficiently use bandwidth.

Thus, the ATM network allocates as much bandwidth as necessary for the call to which access is allowed, but unless the bandwidth of the peak bit rate is provided, the traffic of the connected call is already allocated due to the predictable nature. The bandwidth can be temporarily exceeded.

As a result, the network may be congested. In particular, bursty traffic such as moving images or high-speed data may have a large impact on the congestion. Therefore, in order to eliminate the drawbacks of ATM networks, various preservative and reactive network control techniques are needed, one of which is priority control. Network usage efficiency while ensuring the quality of service of all classes fairly by changing the buffering and transmission priority of cells entering the network according to the network status and the quality of service (QoS) of each class at regular intervals. To increase the

In addition, the method of storing and managing the arriving cell in the queue and the scheduling method of determining the transmission order of the stored cells are collectively referred to as a queue service method, which maintains the quality of service at a level required by the user. Traffic control method for efficient use of resources. Therefore, a new queue service scheme has been proposed according to the characteristics and performance of a queue service scheme suitable for ATM networks.

On the other hand, the queue service method developed around the data communication network is not suitable for ATM networks. There will be many unpredictable services in the ATM network, and the services will have significantly different traffic characteristics, both qualitatively and quantitatively, which is a completely different queue than in a data network with only similar data traffic. It is an important factor in requesting the service method.

In addition, since the service of the ATM network requires a variety of quality of service, it is necessary to handle it differently depending on whether it is a cell loss sensitive service or a delay sensitive service. In the existing data communication network, the loss rate was a major service quality parameter, but real-time services such as voice and video services are included in the ATM network, and as these services become main services, quality of service related to delay becomes important. .

In the case of a data service, the average delay time is a condition regarding a delay, whereas in the case of a real-time service, if the delay condition is very strict and the data is not transmitted within the time limit, it has the same effect as a loss. As such, conditions for varying quality of service and strict delays are another factor requiring the new ATM queuing scheme. As described above, the necessity of a new queue service method suitable for an ATM network has been described, and the requirements for the new queue service method are described in detail.

First, it is necessary to be able to satisfy the requirements for various quality of service, and to optimally handle the service requiring different quality is a method of efficiently using network resources. For example, when the voice and data services are processed identically, the voice service is known to allow only about 250 ms of delay due to the severe condition for delay, but up to 5% of the cell loss, whereas the data service has a cell loss rate. It should be less than 10 ^-10 but there is no big constraint on delay.

In order to satisfy the same quality of service while simultaneously processing the same, they have to guarantee 250ms of delay and 10 ^-10 cell loss. In this case, very large bandwidth and buffer size are required, and the network efficiency is not good. Therefore, the ATM queuing method should be able to enable optimal performance and minimum resource usage through processing appropriate for each service.

In addition, limited delay and cell delay variation must be guaranteed for the real-time service, and the above-described real-time service requires a strict management of the delay since the delay over the time limit is the same as the cell loss. And as important as the limitation on delay, cell delay variation is because irregular queuing delay can cause network congestion and degrade service quality. In particular, in view of efficient management of the network, the function of smoothing the cells is not important.

In addition, there is a need to protect other users from wrong users, and some users send more traffic to the network than they promised when setting up a call intentionally or by a system error. Other users' traffic can be nourished. To prevent this, it is safest to have a buffer of the appropriate size for each connection. When sharing one buffer, the input must be controlled according to the traffic negotiation value.

In addition, the bandwidth should be flexible, and in case of RR (round robin), a bandwidth allocated to one slot is determined according to the length of the frame, and the length of the frame is related to the maximum delay. The shortcoming is that at least the allocable bandwidth becomes quite large. For example, if an RR with a frame size of 10 is used for a link of 100 Mbps, the bandwidth must be allocated in units of 10 Mbps, so the efficiency is not very good.

On the other hand, the algorithm of the stop-and-go queue service method is a queue service method that can be applied to a switching node in a network for real-time traffic-free transmission. In addition, in the stop-and-go queue service method, when real-time traffic enters the network, smoothing can ensure smooth traffic anywhere in the network, providing congestion-free transmission, maximum delay, and cell delay variation. It can be done not to occur at all.

In addition, when describing the stop-and-go queue service method, the concept of a frame is needed first, which can be thought of as a box that continuously delivers packets through a path in a network in order to limit end-to-end delay and cell delay variation. Specifically, for each link, the frame is a continuous interval of time with a fixed period T, and can be seen as moving with the packet between the end points of the link.

1 is a diagram illustrating a composite frame configuration of an output link in a stop-and-go queue service method in a general ATM network. In the case where the size of a frame is fixed, various delay limits are provided. It is difficult and the bandwidth allocation characteristic is lowered. Therefore, Fig. 1 shows the configuration of various types of composite frames in the output link, where the first type frame is a reference frame, and the reference frame is divided into three, and the second type frame is divided into two. The division of the type frame into two is the third type frame.

In this way, the frames of the output link are composed of various types to improve various delay limits and bandwidth allocation characteristics.However, since scheduling of each frame unit gives priority to fixed frames in small order, There was a problem that the use efficiency is lowered.

Accordingly, the present invention is to solve the above-mentioned problems, in order to ensure the end-to-end delay and delay jitter, the output link complex in the stop-and-go queue service method to deliver the packet through the path in the network The frame size is composed of integer multiples of the reference frame, which improves various delay limits and bandwidth allocation characteristics, resulting in better bandwidth density and increased bandwidth usage, thus allowing more calls. A stop-and-go queue service method.

The present invention for achieving the above object, the size of the first type frame in the output link, the time interval of the second type frame k2T1, the time interval of the third type frame k3T1, After setting the time interval of the four-type frame to k4T1 and the time interval of the n-type frame to knT1 (where 1 <k2 <k3 <k4 <… n), EDD is determined based on the end point of each frame. It is characterized by consisting of a method for selecting a cell to serve by applying.

According to the present invention configured as described above, the size of a frame is an integer multiple of a reference frame in a stop-and-go queue service method in which packets are mounted on a frame and delivered through a path in a network to ensure end-to-end delay and delay jitter. Various delay limits and bandwidth allocation characteristics are improved, resulting in better bandwidth density and increased bandwidth utilization to accommodate more calls.

1 is a diagram illustrating a composite frame configuration of an output link in a stop-and-go queue service method in a general ATM network;

FIG. 2A illustrates a switching node n having an output link l connected to an input link l ', l' 'in a typical stop-and-go queue service scheme. FIG.

FIG. 2B is a diagram showing the structure of a frame in each link shown in FIG. 2A;

3 is a diagram illustrating a composite frame configuration of an output link in a stop-and-go queue service method using EDD in an ATM network according to the present invention;

FIG. 4 is a diagram illustrating a priority order among composite frames for each reference frame interval in the output link shown in FIG. 3.

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in detail.

On the other hand, the delay of the ATM cell can be divided into two components, a constant size delay and an irregular delay. In the former case, the delay caused by the cell disassembly and assembly of the AAL layer is the main cause. When the ATM network and the existing PSTN network are interworked, it causes echo problems, and the latter causes delays in the multiplexing or switching process in the ATM layer, and transmission of operation, administration and maintenance (OAM) cells in the physical layer. This is caused by delay.

In addition, traffic having various quality of service and traffic characteristics is multiplexed and transmitted in an ATM network, and traffic having a variable bit rate allocated to a bandwidth below a maximum bit rate may exceed the negotiated bandwidth, which causes congestion. Therefore, priority control is performed to guarantee quality of service of a class that is sensitive to time delay through cell loss distribution and cell transmission order control through selective cell storage, discarding, and transmission amount control.

In addition, priority control is used to control the priority according to the quality of service between different traffic classes, and the low priority indicated by the priority bits of the cell head within the same traffic such as MPEG video. And control the storage and transmission of high priority cells.

In addition, the queue service discipline refers to a method of storing and managing an arriving cell in a queue and a method of determining a transmission order of stored cells, which guarantees user service quality and improves network utilization efficiency. Traffic control method to maximize. The queue service method is necessary wherever there is a queue such as B-NT (BISND-Network Termination) at the network entrance or ATM switch across the network, and the traffic arriving at the B-NT is irregular compared to the traffic in the network. Therefore, the queue service method in B-NT has a very important effect on network efficiency and service quality.

In addition, the queue service method has been mainly developed based on the existing data communication network, and the transmission delay variation is not a concern, and an attempt has been made to reduce the average transmission delay by serving a short packet first. First come first serve (FCFS), which serves packets that arrive first, RR (round robin) that provides fair service by providing separate buffers for each channel, and head of line (HOL) that serves strict priorities. It is a representative queue service method of the existing data network.

However, in order to efficiently accommodate services of various characteristics in a BISDN environment where services are integrated and speeded up, research on a completely different queue service method is required. Therefore, consideration of cell delay variation (CDV) is essential due to the integration of real-time services, and the quality of each service must be guaranteed for each service, and the quality of other services is maintained even when there is a connection generating excessive traffic. You should be able to.

FIG. 2A is a diagram illustrating a switching node n having an output link 1 connected to an input link 1 ′, 1 ″ in a typical stop-and-go queue service scheme, and FIG. 2B is shown in FIG. It is a figure which shows the structure of the frame in each link.

First, the most central idea of dealing with frames is that packets arriving during frame F 'input to input link l' must all be sent during frame F of the output link for transmission to output link l. . Here, frame F refers to a frame that is not overlapped for the first time after the end of frame F '. This frame F is called a frame adjacent to frame F'.

On the other hand, if a packet of the frame F 'cannot find an empty slot during the frame F, it is discarded without waiting for an empty slot to be generated in the next frame. Therefore, in order to prevent the packet from being discarded, all packets input from the input link must be able to find an empty slot in an adjacent frame of the output link, and an appropriate connection control function is required for this purpose.

In addition, the stop-and-go queue service method can be implemented by slightly modifying a conventional FIFO queue, and a packet arriving for a predetermined frame must be stored once for transmission to an adjacent frame, and serviced only for an adjacent frame. Because only a FIFO queue and a simple service controller are required for each link.

In order to improve the efficiency of the network, it is possible to support connection and delay-sensitive data, that is, best effort (BE) traffic, which allows a certain amount of loss. After transmission, the remaining packets are then serviced according to the desired loss rate, and after that, the service is only served if there is an empty slot in the frame.

On the other hand, because the packets are continuously transmitted through adjacent frames, the end-to-end delay experienced by one packet is almost constant. That is, Dp = D + d, where d denotes a cell delay variation, which occurs because of positions occupied by packets in the input frame and the output frame, and satisfy -T ≦ d ≦ T.

In addition, the constant D can be expressed by the formula D = τ + HT, where τ is the total transmission delay and H is the number of links passed. Since the T is the size of the frame, it is also possible to adjust the delay. As described above, if the signal source performs the connection acceptance control and the stop-and-go queue service method in the network, no loss occurs at all while ensuring maximum delay and maximum cell delay variation. It can also have the characteristics. Therefore, it is effective for the transmission of real-time traffic in a high speed packet network.

3 is a diagram illustrating a composite frame configuration of an output link in a stop-and-go queue service method using EDD in an ATM network according to the present invention. The configuration shows that the first type frame is a reference frame, wherein twice the reference frame is a second type frame, three times the reference frame is a third type frame, and four times the reference frame is a second frame. It becomes four type frames.

First, the stop-and-go algorithm delivers packets over frames in a network to guarantee end-to-end delays and delay jitter, and the frames have a continuous time with a fixed period (T) for each link. It is the interval between and it can be considered that the end-to-end of the link moves with the packet.

And, the most basic thing dealing with the frame is that the packets arriving during the frame F 'input to the input link l' are all output during the frame F of the output link for transmission to the output link l, and the frame ( If the packet of F ') is not output during the frame F, it is discarded without waiting for the next frame. Therefore, all packets input from the input link must be output during the corresponding frame of the output link, which requires proper connection acceptance control.

On the other hand, when the frame size, i.e., the time interval is fixed to T, it is difficult to provide various delay limits, thereby degrading bandwidth allocation characteristics. Therefore, in the stop-and-go algorithm, the size of the frame is T1 = kT, T2 = kT1, T3 = mT2, T4 = nT4,... Various delay limits and bandwidth allocation characteristics can be improved by using various types, but the bandwidth usage decreases because the frame size is given high priority in the order of each frame, for example, maximum bandwidth usage. The efficiency is limited to about 70%.

Therefore, without using the fixed priority in order of the size of the frame, by first serving the cells of the frame closest to the deadline based on the end of each frame it is possible to increase the bandwidth usage efficiency.

FIG. 4 is a view showing a priority order among composite frames for each reference frame interval in the output link shown in FIG. 3. First, a frame preferentially selected in a time interval, for example, T1, may be a first type frame or a first frame. The service is selected in the order of the type 2 frame and the type 3 frame.

In the time interval T2, the first type frame and the second type frame are selected and serviced in the same order, and then the third type frame and the fourth type frame are selected and serviced. In addition, in the time interval T3, the first type frame and the third type frame are selected and serviced in the same order and preferentially, and then the second type frame and the fourth type frame are selected and serviced.

In the time interval T4, the first type frame and the second type frame are selected and serviced in the same order, and then the fourth type frame and the third type frame are selected and serviced. Further, in the time interval T5, the first type frame and the fourth type frame are selected and serviced in the same order and preferentially, and the second type frame and the third type frame are selected and served in the same order.

On the other hand, the reference numerals written in the components of the claims of the present application to facilitate the understanding of the present invention, and are not written in the intention to limit the technical scope of the present invention to the embodiments shown in the drawings.

As described above, according to the present invention, in the stop-and-go queue service method in which a packet is mounted in a frame and delivered through a path in a network in order to ensure end-to-end delay and delay jitter, the size of the frame is equal to that of the reference frame. It is composed of several kinds of integer multiples, so that various delay limits and bandwidth allocation characteristics are improved, resulting in better bandwidth density characteristics and increased bandwidth usage to accommodate more calls.

Claims (3)

In the stop-and-go queue service method in which packets are mounted in a frame and delivered through a path in a network to ensure end-to-end delay and delay jitter, the size of the first type frame in the output link is set at a time interval T1. The time interval T2 of the second type frame is k2T1, the time interval T3 of the third type frame is k3T1, the time interval T4 of the fourth type frame is k4T1, and the time interval is nth type frame. After setting (Tn) to knT1 (where 1 <k2 <k3 <k4 <… n), a method of selecting a cell to be served by applying EDD on the basis of the end of each frame is characterized by the above-mentioned. A stop-and-go queue service method using EDD in an ATM network. The method of claim 1, wherein the time interval T1 of the first type frame is set as a reference frame. 2. The EDD of claim 1, wherein the time intervals T2 to Tn of the second type frames to the nth type frames are formed based on the time interval T1 of the first type frames. Stop-and-Go Queue Service Method