JPH0652119A - Replication file referencing system - Google Patents

Abstract

PURPOSE:To provide a system for efficiently using a replica server instead of a file server in a distributed data processing system. CONSTITUTION:A client 120 connected to a communication network 152 gives the reference request of a file 112 in a file server 110 through a communication network 151. When a replica 122 is stored in the disk of the client 120, a gate 140 in the mid-way of a path recognizes it and pieces of information specifying 110, 112, 120 and 122 are written in an update fact table 143. When a client 130 refers to the file 112, the gate 140 in the mid-way of the path recognizes the reference request and alters the reference request to be transmitted to the file server 110 in an original case so that it is to be transmitted to the client 120. The client 120 plays the role of the replica server only to the file 112.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a replicated file reference method suitable for efficient operation of replicated files in a distributed data processing system.

[0002]

2. Description of the Related Art A distributed data processing system is a LAN (L
This is a system in which a plurality of data processing systems such as personal computers and workstations are connected to a communication network called "Ocal Area Network", and each process is performed while exchanging information with each other via the communication network. As an important technique in the distributed data processing system, there is a distributed file system which enables one data processing system on a communication network to access a file existing in another data processing system. Distributed file system is 1
It is possible to share a file existing in one data processing system by a plurality of data processing systems. Here, a data processing system in which a shared file exists is called a file server, and a data processing system that accesses a shared file on the file server is called a client.

When a specific shared file is frequently accessed, requests are concentrated on the file server, which may become a bottleneck in the performance of the entire distributed data processing system. Therefore, a copy of this file is also placed on the second file server, and a part of the access request to this file is processed by the second file server, whereby the load is distributed between the two file servers, and the bottle The neck is eliminated. By placing copies of the same file on multiple data processing systems, by using the copy as an alternative in the event of file corruption,
Since access to the file is not interrupted and recovery is easy, the reliability of the entire distributed data processing system is improved. Each copy of the same file is called a replica of that file, or a replicated file,
Creating a copy is called replication, and a data processing system that owns a copy of a file is called a replica server. Regarding distributed file system, ACM, Computing Surveys, Vol. 22, No. 4 (1990) PP321-374 (A
CM, Computing Surveys, Vol.
22, No. 4 (1990) PP321-374), and replication is discussed in PP339-340 of the same document.

[0004]

The distributed file system is expected to be operated not in a single LAN but in a worldwide network in which LANs are interconnected.
For example, it is used to refer to a file in an office in the United States from an office in Tokyo or Osaka. At that time, the use of the above-mentioned replication is important for suppressing the concentration of the overhead of the file server and the increase in the communication volume on a global scale. In the above example, consider a method in which clients in Tokyo and Osaka access the US twice, and a method in which a US replica server is placed in Tokyo and a client in Japan accesses the replica server.
Compared to the former, the latter generally takes less time to communicate and does not have the overhead of US file servers.

However, where to put the replica server is actually a problem. Conventionally, for file servers that exist in a long distance (US), a dedicated server for replication has been installed. The client needed to refer to the replica with steps that were aware of the location of the dedicated server. Further, the concentration of access to the replica server by a plurality of clients inevitably increases the processing overhead of the replica server and increases the traffic of the communication network. An object of the present invention is to make a client that first refers to a file in a distant file server a replica server of the file, and when another client refers to the file in the file server, the replica server instead of the file server. It is to provide a replicated file reference method that allows you to refer to a replica in the.

[0006]

In order to achieve the above object, the present invention provides an update fact table in a gate for controlling communication between communication networks, and the update fact table includes data processing in its own communication network. The data processing device (replica server) in which the file server and the file and the replicated file exist, which indicates that the device stores the replicated file in the file storage means by referring to the file existing in the file server in the other communication network, and its replicate Information that identifies a file is registered. The gate monitors communication from a data processing device in its own communication network to a data processing device in another communication network, and when the communication refers to a file existing in a file server, updates the update fact table. It is checked to see if the file server and file are registered. Then, when not registered, the communication with the file server is permitted, the data processing device of the communication source with the file server and the file is recognized from the communication result, and the file server, the communication source with the file and the communication source are recognized in the update fact table. Information that identifies a certain data processing device (replica server) and its replicated file is registered.

When the file is registered, the reference destination of the file is changed from the file server to the replica server in which the replicated file of the file exists, so that the data processing device of the communication source can refer to the replicated file.

[0008]

When a replica server that stores a replicated file of a file that exists in a file server for which a client requests a reference is present, the client requests a reference request from the file server to the replica server, and therefore communicates with the file server. Communication is much faster, and efficient file reference is possible.
In addition, by not having a specific dedicated replica server, the concentration of network traffic and the load on the server are reduced. Further, by adding information on the expiration date of the replicate file, it is possible to easily refer to the optimum replicate file when there are a plurality of replicate files of the same file.

[0009]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an outline of a distributed data processing system to which the present invention can be applied. File server 11 which is a data processing system in which shared files exist
0, clients 120 and 13 which are data processing systems for accessing shared files on the file server
0 is an independent data processing system, each data processing system is connected to a common communication network 150 and 152, and the gate 140 and the gate 14 are connected.
The two can communicate with each other and transfer data via the wide area network 151. Each gate comprises a data processing system, and a communication fact table 143 is connected to the gate 140. For example, the file server 110, the clients 120, and 130 are high-performance workstations (WS), the gate 140 is a switch capable of high-speed memory access, and there is an update fact table 143 on the memory, and communication is performed. Using a local area network (LAN) as the network is an example of a typical configuration. The file server 110 and the client 120 respectively have disk devices 111 and 121 for storing files, and the files 112, 113, and 122 stored therein are the client 1
It can be referred to from 30.

Consider a particular file 112. At first,
Client 1 refers to file 112 to server 110
20 demands. The server 110 allows the client 120 to create a replica file. The client 120 stores a copy of the file 112 in the disk device 121 as a replica file 122.
A client in which a replica file exists is called a replica server. The client 120 in which the replica file 122 exists becomes the replica server. During this series of operations, the data that allows the replica passes through the gate 140. The replica permission data is added to the main body of the copy of the file 112 or is sent separately from the server 110 to the client 120 before copying. The gate 140 monitors the replica permission data transferred on the network, and when it finds the replica permission data, the file 1 of the server 110 is detected.
The twelve replicas store the update facts created under the control of the client 120 in their own table 143.

After that, the server 1 refers to the file 112.
10 when the client 130 makes a request to the gate 14
0 recognizes the fact of demand for transmitting a communication network,
The update fact table 143 is searched. When the reference destination indicated by the request fact matches the update destination indicated by the update fact, the reference destination of the client 130 is changed from the server 110 to the replica server 120. Therefore, the client 13
The file reference request of 0 is processed by the replica server 120. The client 130 will access the replica 122 instead of the file 112.

FIG. 2 is a conventional configuration diagram. The existence of the dedicated replica server 220 is a big difference from FIG. In FIG. 2, the client 230 and the replica server 220 are on the same communication network 252, but in many cases they actually exist in different networks, and the client 230 often does not know the existence of the replica server 220. . Therefore, client 2
The 30 goes to the file 211 in the file server 120 without knowing the replica 221 in the replica server 220.

The present embodiment is characterized in that the client becomes a replica server for individual files without providing a dedicated replica server. In FIG.
Another file 113 is the client disk 131
Stored as a replica 133 in the. Therefore, clients 120 and 130 are
12 replica servers and file 113 replica servers. The clients 120 and 130 need only perform the process of requesting the file reference to the file server 110, and the gate 140 takes charge of the process of changing the file reference destination to the replica server when the request passes. The communication via the communication network is carried out by taking over a protocol that defines the communication procedure of the transmitting side and the receiving side, and this also includes the method of specifying the data processing device to be received, that is, the method of specifying the address. The data transfer itself is performed by generating or detecting electric vibration on the communication path.Communication is performed by each data processing system by reading the address information included in the detected electric vibration and communicating with itself. It is established by determining whether or not there is.

FIG. 3 shows the structure of the gate 140. The conventional gate 240 is the communication network 25.
It only has a data delivery means 300 for delivering data between 0 and 252 to each other. The gate 140 of the present embodiment is independent of the data delivery means 300, the reference destination changing means 301,
It has an update fact recognition means 302, a reference request recognition means 303, a communication monitoring means 304, an update fact table 143, and connectors 310 and 311. Means 300 to 304 are
It is a separate processor and has a multiprocessor configuration, and each processor operates under a separate program. It is desirable that the update fact table 143 is data existing on a memory commonly accessed by the means 301 to 303 in order to speed up data processing. Needless to say, the means 300 to 303 may be configured by one processor instead of separate processors, or may be configured by a small number of multiprocessors.

FIG. 4 shows an example of the update fact table 143. A column 401 showing an update source (server), a column 402 showing an update destination (client), and a column 403 showing an expiration date designated by the update destination. For example, column 4
The data in 01 and the field 402 are character strings that combine the network address and the file name of the data processing system, and the data in the field 403 is an integer indicating the world standard time.

FIG. 5 shows a processing procedure when the gate 140 receives replica update communication data. Step 50
1 to 503 are procedures for delivering communication data, and steps 510 to 511 are procedures for recognizing update information. FIG. 6 shows a processing procedure when the gate 140 receives file-referenced communication data. Steps 601-04
Is a procedure for delivering communication data, and steps 610 to
615 is a procedure for changing the reference destination of communication data.

The operation of the gate 140 according to this embodiment will be described below with reference to FIGS. 1 and 3 to 6. First, the client 120 sends the file 11 to the server 110.
The operation when referring to 2 will be described. The connector 310, which is an interface with the communication network 152, processes the destination address of communication in addition to electrical connection. That is, when the designated address is the address toward the network 150, the communication data is fetched and the processing is performed by the data delivery means 300 (step 601). The data delivery means 300 matches the communication data with the protocol of the wide area network 151 (step 602), and the connector 31
Communication data is transmitted from 1 (step 603).

The communication data of the file 112 and the data indicating the permission of the replica are returned to the data delivery means 300 via the connector 311 (step 501). The data delivery means 300 sends the communication data from the connector 310 (step 503) according to the protocol of the communication network 152 (step 502). At that time, the communication monitoring means 304 monitors the signal delivered by the connector 311 to the data delivery means 300 (step 51).
0), A signal is transmitted to the update fact recognition means 302. The update fact recognizing means 302, when it recognizes that the communication data is data to be updated by the replica, in a format as shown in FIG.
The update fact is stored in the update fact table 143 (step 511).

Next, the client 130 makes the server 11
The operation of referencing the file 112 to 0 will be described.
Connector 310, data delivery means 300, connector 31
The reference request passes through the gate 140 via 1 (steps 601 to 603). The communication monitoring means 304 monitors the signal delivered by the connector 310 to the data delivery means 300, and sends the signal to the reference request recognition means 303 (step 610). When the reference request recognition unit 303 interprets the signal and knows that it is a file reference request to the file server, it searches the update fact table 143 (step 6).
11) Check if the same file already exists as a replica in the communication network 152. If the update fact does not exist in the update fact table 143, the processing ends. If the update fact exists in the update fact table 143 (step 612), the reference request recognition means 3
03 interrupts the data delivery means 300 by interrupt 615 (step 613), and stops the data delivery means 300 from delivering the file reference request to the communication network 151 via the connector 311 (step 604).
After the cancellation, the reference request recognition means 303 is the reference destination change means 3
Pass control to 01. The reference destination changing unit 301 changes the data reference destination based on the update facts in the update fact table 143 (step 614). The data delivery means 300 delivers the reference request to the change destination replica server (step 616), and the change destination replica server 120 delivers the reference data to the client 130 (step 61).
7).

Next, using the embodiment of the transmission data 700 from the client 130 to the file server 110 shown in FIG. 7 and the reception data 710 obtained as a result, steps 614, 616 and 617 shown in FIG. The change of the reference destination of the data and the data transfer after the change will be described. The reference request 700 is a source address 701.
And a data section 703 indicating a reference request of a specific file 111 and a header for communication such as a destination address 702.
And a flag 704 (eg, a code in a bit string) that allows the gate 140 with the update fact table 143 to change the recipient. The flag 705 is the latest replica reference flag (this will be described later). The communication monitoring unit 304 uses the reference request flag 704.
Is set, the subsequent processing for referring to the replica server 120 instead of the file server 110 is determined. When the processing for referring to the replica server 120 (steps 611 to 614) is performed and the reference destination is changed, the reference destination changing unit 301.
Overwrites the data of the destination address indicating the replica server 120 on the destination address 701 designated by the client 130. The data delivery unit 300 delivers the reference request to the change destination replica server 120 instead of the file server 110. Result 710 of the reference request 700 that the replica server 120 delivers to the client 130.
Is the address 711 of the client 130, the address 712 of the replica server 120, the data section 713 indicating the contents of the file server, and the flag 7 indicating that the data is from the replica server 120 instead of the file server.
Including 14.

In another embodiment of step 614 of FIG. 6 of the present invention, an update fact table 143 as shown in FIG. 9 is provided,
As shown in FIG. 8, when a plurality of replicated files of the same file exist, refer to the replica last passed through the gate 140 (latest replica) or the replica with the longest reference expiration date (longest valid replica). Provide instructions to do so. The update fact table 143 of FIG. 9 has an original file name 901 that combines a server name and a file name, a file name 902 that combines a replica server name and a replicated file name, and the validity of the replica notified from the file server. An example is shown in which a deadline 903 and a time 904 when the data of the items 901 to 903 are updated when passing through the gate are included. In FIG. 8, the reference destination changing unit 301 causes the latest replica reference flag 70 included in the reference request 700 transmitted by the client 130.
It is determined whether 5 (for example, a bit string) is set (step 801). Reference request file 12
When the latest reference flag 705 is set when the file name 703 matching 1 exists, the file update time 904 is searched, and the replicate file 902 having the latest update time is set as the change destination (step 803), and set. If not, the expiration date 903 of the replica is searched, and the replicate file 902 having the longest reference expiration date is used as the change destination (step 802).

[0022]

According to the present invention, in a network interconnected on a global scale, an update fact table is provided at a gate in the middle of passage, and by using this, the client does not refer to the file in the file server but , The replicated file in the replica server can be referenced, which has the effect of enabling high-speed file browsing. Further, by not providing a specific fixed dedicated replica server, it is possible to avoid the concentration of traffic, promote the load distribution of the distributed file system, and reduce the load on the server. Furthermore, by adding the expiration date information of the replicate file, it is possible to easily refer to the optimum replicate file when a plurality of replicate files of the same file exist.

[Brief description of drawings]

FIG. 1 is a block diagram showing an outline of a distributed data processing system to which the present invention is applied in one embodiment.

FIG. 2 is a block diagram showing an outline of a conventional distributed data processing system.

FIG. 3 is a block diagram showing a configuration of a gate.

FIG. 4 is a diagram showing a structure of an update fact table.

FIG. 5 is a flowchart showing processing in the gate when the gate receives replica update communication data.

FIG. 6 is a flowchart showing a process of changing a reference destination in a gate.

FIG. 7 is a diagram showing transmission data from a client to a file server and reception data obtained as a result.

FIG. 8 is a flowchart of replica reference processing when a plurality of replicated files of the same file exist.

FIG. 9 is a diagram showing an example of an update fact table.

[Explanation of symbols]

 110 file server 111, 121, 131 disk 112, 113 file 120, 130 client 122, 133 replica file 140 gate 143 update fact table 150, 152 network 151 wide area network 300 data delivery means 301 reference change means 302 update fact recognition means 303 Reference request recognition means 304 Communication monitoring means 310, 311 Connector

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Hirohiko Nakano Inventor Hirohiko Nakano 1099, Ozenji, Aso-ku, Kawasaki-shi, Kanagawa Inside the Hitachi, Ltd. System Development Laboratory (72) Inventor Atsushi Kobayashi 5030 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Ceremony Company Hitachi Software Development Division

Claims (5)

[Claims] 1. A distributed communication system comprising a plurality of communication networks, each communication network having a gate for controlling communication between the plurality of data processing devices having file storage means and the communication networks, and enabling communication between the data processing devices. A replicated file reference method in a data processing system, wherein the gate stores a replicated file in a file storage unit by a data processing device in its own communication network referring to a file existing in a file server in another communication network. The data processing device (replica server) in which the file server and the file and the replicated file exist, and the update fact table in which the information identifying the replicated file is registered, other communication from the data processing device in the own communication network network Monitoring the communication to the data processing device in the inside, and when the communication refers to a file existing in the file server, refers to the update fact table to check whether the file server and the file are registered or not. When not registered, the communication with the file server is permitted, the data processing device of the communication source with the file server and the file is recognized from the communication result, and the file server is the communication source with the file in the update fact table. Information that identifies the data processing device (replica server) and its replicated file is registered. When registered, the file reference destination is changed from the file server to the replica server in which the replicated file of the file exists, and Verify that the data processing unit is configured to reference the replicated file. Replicated file reference method for the butterflies. 2. The replicated file reference system according to claim 1, wherein the data processing device uses the gate to change the receiving destination from the file server to the replica server during transmission data when communicating with the file server. The replicated file reference method is characterized in that a destination change permission flag is provided for permitting it. 3. The replicated file reference method according to claim 1, wherein the file server adds the replicated file permission data to the file information of the communication result in response to the file reference request from the data processing device, and performs the data processing. The device creates a replicated file based on the permission data of the replicated file, the gate detects the permission data of the replicated file by monitoring the communication on the communication network, and the update fact table based on the permission data. A replicated file reference method characterized by updating the file. 4. The replicated file reference method according to claim 3, wherein the file server adds data indicating a valid reference time limit to the permission data of the replicated file, and the data processing device is used when communicating with the file server. The latest replica reference flag indicating the latest replicate file reference request is provided in the transmission data, and the gate registers the update time together with the valid reference time limit when updating the update fact table based on the permission data, and the data processing When the latest replica reference flag is set in the transmission data of the device, the reference destination is changed to the replica server in which the latest replicated file exists, and when the latest replica reference flag is not set, the valid reference time limit is the most. Replica with long replicated files A replicated file reference method characterized in that the reference destination is changed to the server. 5. The replicated file reference method according to claim 1, wherein the gate includes a data delivery unit that controls communication between a data processing device in the own communication network and a data processing device in another communication network, and the communication. Communication monitoring means for monitoring the contents of the above, an update fact recognition means for recognizing that the communication data sent from the communication monitoring means relates to a replicate file, and updating the update fact table, and the communication monitoring means Recognizing that the transmitted communication data is a file reference request to the file server, interrupts the data delivery means to stop the delivery of the communication data to other communication networks, and passes control to the reference destination changing means described later. Reference request recognition means, and referring to the update fact table based on the communication data of the file A replicated file reference system comprising a reference destination changing unit for changing a reference destination and instructing the data delivery unit to perform the change.