CN101217571B - Methods for write/read file operations in a multi-replica data grid system - Google Patents

Abstract

The method for writing/reading file operations in a multi-copy data grid system belongs to the field of high-speed file writing operations in a multi-copy data grid system. ; After successful identity authentication by the metadata server, select a replica storage server that is logically closest to the user through the server distance comparison module, and select the corresponding replica view information, and confirm it through the view analysis module of the replica access client After the replica view is valid, the read-write execution module of the replica access client sends a write data request to the replica storage server, and when the replica storage server executes the write operation successfully, the replica view updating module of the metadata server updates the replica view . The invention enables the user to select the closest copy storage server to perform the write operation, and at the same time ensures that the user can normally perform the read operation.

Description

Methods for write/read file operations in a multi-replica data grid system

technical field

The invention relates to the file operation in the data grid system, in particular to the high-speed write file operation in the multi-copy data grid system.

Background technique

Data grid systems widely use data copy technology to improve the performance of file operations, that is, multiple file copies are stored in different data nodes, including a rewritable master copy file and multiple read-only copy files. For the file read operation, in general, the user can manually select the nearest copy file for the read operation, or the resource scheduler in the data grid automatically assigns the user a copy file with low communication overhead for the read operation. For file writing operations, data is generally written to the master copy file. The master copy file is the only rewritable copy file. After the master copy file is updated, other read-only copy files will be automatically updated synchronously or asynchronously.

It can be seen from the above prior art that the current file writing operation in the data grid has the following deficiencies:

Since only the main copy file can be rewritten, when the transmission overhead of storing the main copy file to the user is high, the performance of the data grid's file writing operation will be reduced.

When frequent write operations are performed on a file, the data node storing the master copy file may be overloaded, resulting in a sharp drop in the performance of file write operations, or even a system crash.

Contents of the invention

The purpose of the present invention is to provide a method for nearby writing/reading of duplicate files in a multi-duplicate data grid system, which enables users to write operations on the latest duplicate files and improves the performance of file writing operations.

A method for supporting nearby writing/reading of replica files in a multi-replica data grid system involves a metadata server, a replica storage server and a replica access client.

The metadata server includes a copy view storage module, a copy view update module, a server distance comparison module, a copy view selection module and an error handling module. The functions of each module are as follows:

Copy view storage module: store copy view information, describe the complete view information of a copy by the address of the copy storage server and the data segment description information of the available copies stored on the copy storage server, the data segment description information is composed of The starting offset and the length of the data segment in , the address of the copy storage server is represented by the domain name or IP address of the copy storage server;

Replica view update module: according to the specific situation of replica update, find and delete outdated replica view information, create new replica view information, and modify the data segment description information of replica view;

Server distance comparison module: Estimate the logical distance between each copy storage server and the user who initiates the read and write request. The farther the distance, the smaller the available bandwidth of the communication link between the server and the user, and vice versa;

Copy view selection module: select the copy view with the smallest distance from the user according to the user's read and write request and the output result of the server distance comparison module; when the user performs a file read operation, the copy view with the smallest distance may be selected by Multiple data segments on multiple replica storage servers are stitched together; when a user performs a file write operation, the replica view with the smallest distance is a data segment on the replica storage server closest to the user;

The replica storage server is responsible for storing and managing replica files; responsible for receiving and executing read and write operations from users; responsible for reporting the update status of the managed replica files to the metadata server;

Error handling module: when identity authentication fails, record an exception and send a request failure response to the client.

The access client includes a user access interface module, a view request module, a view analysis module, a read-write execution module and an error handling module. The functions of each module are as follows:

User access interface module: receive read and write file requests from users;

View request module: according to the output parameters of the user access interface module, send read and write requests to the metadata server to obtain the copy view information required for specific read and write operations;

Replica view parsing module: parse the replica view information returned by the metadata server, group the replica view information according to the servers where the data segments in the replica view information are located, and the data segments stored on the same server belong to the same group, and the same group Each data segment is sorted within the group according to the starting offset; for read operations, the read and write execution module obtains the data of each data segment from the relevant replica storage server, and forms a complete data segment and returns it to the user access interface module; for write Operation, write user data into the corresponding data segment of the corresponding replica storage server through the read-write execution module;

Read and write execution module: interact with the replica storage server to perform read and write file operations;

Error handling module: handle the errors generated when reading and writing view requests and executing read and write files.

Among them, the method of writing files by accessing the client, the implementation steps are as follows:

Step (1). The user submits a file writing request to the user access interface module of the copy access client in the system,

The request includes: the identifier of the file to be written, the data to be written, the offset of the starting address of the written data segment and the length of the data segment;

Step (2). After the user access interface module simply encapsulates the file writing request, it sends it to the view request module of the copy access client, and the view request module authenticates the identity of the metadata server in the system. The module issues a write view request;

Step (3). The identity authentication module of the metadata server performs simple authentication on the write view request. If it is not authenticated and passes the authentication, the error message is sent to the error processing module in the metadata server, and the exception is recorded. Afterwards, send a write view request invalidation response to the view parsing module in the copy access client in the system; if: pass the authentication, then send the authenticated correct information and the user’s write view request to the server distance in the metadata server A comparison module, the server distance comparison module estimates and selects a copy storage server with the smallest logical distance from the user who initiates the write view request according to the domain name or IP address of each copy storage server stored in advance;

Step (4). The copy view selection module in the metadata server according to the copy storage server and the user write view request obtained from the server distance comparison module, from the copy view storage module in the metadata server. In the copy view, select the copy view, generate corresponding copy view information, and return the result to the view parsing module in the copy access client in the system;

Step (5). After the view parsing module of the copy access client receives the copy view information returned from the copy view selection module in the metadata server or the write view request failure response returned by the error handling module, check whether it is Valid view information; if it is an invalid view, send the error message to the error handling module of the copy access client, and return the error message to the user through the user access interface module; if it is a valid view, then analyze the copy view information , group the replica view information according to the servers where the data segments in the replica view information are located, so that the data segments stored on the same server belong to the same group, and then group the data segments of the same group according to the offset of the starting address The size of the group is sorted, and the group view information is sent to the read-write execution module in the copy access client to perform specific write data operations; step (6). The read-write execution module of the copy access client is based on the The grouped view information received by the view parsing module sends a write data request to the copy storage server;

Step (7). The copy storage server described in the step (6) checks the validity of the write data request received from the read-write execution module in the copy access client; if it is an invalid data request, it sends the request to the copy The read-write execution module of the access client returns an error message; if it is a valid request, the data copy is executed to write data, and after the write is successful, the copy storage server sends a copy update to the copy view update module in the metadata server Report;

Step (8). After the replica view update module of the metadata server in step (7) receives the replica update report obtained from the replica storage server, it updates the corresponding replica view information through the replica view storage module in the metadata server. Perform an update operation to provide an accurate copy view to the copy view selection module, if the copy view update fails, the copy view update module returns copy update failure confirmation information to the copy storage server described in step (7), and then the The copy storage server rolls back to the state before the write operation, and sends a write data failure response to the read-write execution module of the copy access client described in step (6);

Step (9). If the copy view update described in step (8) is successful, the copy view update module described in step (8) returns copy update success confirmation information to the copy storage server described in step (7), and the copy storage The server sends a write data success response to the read-write execution module of the copy access client described in step (6);

Step (10). After the read-write execution module of the replica access client in step (6) receives the write data response from the replica storage server described in step (7), it judges the write data response, and if it is a write data failure response , then send the error information to the error handling module of the copy access client, and return the error information to the user through the user access interface module; if it is a successful response to writing data, then return the success information of writing data to the view parsing module, The view parsing module returns a file writing response to the user through the user access interface module.

Wherein when the method is used for the read operation, after the identity authentication is passed, the server distance comparison module selects and generates corresponding read view information according to the logical distance information between the copy storage and the user.

The user needs to obtain a file handle before performing read and write operations, and release the obtained file handle after completing all read and write operations.

In summary, using the method provided by the present invention can enable the user to select the nearest copy storage server to perform the write operation, and at the same time ensure that the user can normally perform the file read operation and improve the performance of the file read and write operation. Write details are shielded to provide users with more convenient file operation services.

Description of drawings

Fig. 1 is a system topology diagram of the present invention;

Fig. 2 is a block diagram of the main modules of the metadata server;

Fig. 3 is a block diagram of the main modules of the access client;

Fig. 4 is the flow chart that the user carries out writing file operation;

Fig. 5 is a flow chart of the user's file reading operation.

Detailed ways

The implementation steps of the method for writing a file operation by accessing the client in the present invention are as follows:

1. The user submits a file writing request to the access interface module of the access client of the access client, and the request includes the identifier of the file to be written, the data to be written, and the starting offset and length of the data segment to be written;

2. After the access interface module simply encapsulates the file write request, it is handed over to the view request module, which requests the view information required for the write operation from the metadata server;

3. The metadata server performs simple authentication on the request. If the authentication is not passed, an error message will be returned to the access client; if the authentication is passed, it will transfer to the server distance comparison module, estimate and select the copy storage server closest to the user;

4. After finding the nearest copy storage server, the copy view selection module generates the corresponding view information, and returns the result to the access client;

5. The view analysis module of the access client receives the view information returned by the metadata server, and checks whether it is a valid view information. If it is an invalid view, it will return an error message to the user through the access interface module; if it is a valid view, it will follow the view information Try to send a write data request to the nearest replica storage server through the read-write execution module;

6. The replica storage server checks the validity of the data write request from the access client, and if it is an invalid request, returns an error message to the access client; if it is a valid request, executes the write operation; after the write is successful, the replica storage server sends the metadata The server sends a copy update report; if the update fails, it returns an error message to the access client and rolls back to the state before the write operation; if the update is successful, it notifies the access client;

7. If the write request of the view analysis module of the access client fails, an error message is returned to the user through the access interface module; if successful, the user is prompted through the access interface module that the file writing operation is successfully completed.

The implementation steps of the method for reading files by accessing the client are as follows:

1. The user submits a file read request to the access interface module of the access client, and the request includes the identifier of the file to be read, the starting offset and length of the data to be read;

2. After the access interface module simply encapsulates the read file request, it is handed over to the view request module, which requests the view information required for the read operation from the metadata server;

3. The metadata server performs simple authentication on the request. If the authentication is not passed, it will return an error message to the access client; if the authentication is passed, it will transfer to the server distance comparison module, estimate the distance between the replica storage server and the user, and transfer Enter the copy view selection module for processing;

4. The copy view selection module selects and generates corresponding view information according to the distance information obtained by the server distance comparison module, and returns the result to the access client;

5. The view analysis module of the access client receives the view information returned by the metadata server, and checks whether it is a valid view information. If it is an invalid view, it will return an error message to the user through the access interface module; if it is a valid view, it will follow the view information Attempt to send read data requests to one or more replica storage servers through the read-write execution module;

6. The replica storage server checks the validity of the read data request from the access client, and if it is an invalid request, it returns an error message to the access client; if it is a valid request, it executes the read operation and returns the read data to the access client end;

7. If the read request of the view analysis module of the access client fails, an error message will be returned to the user through the access interface module; returned to the user.

The specific implementation of the present invention will be further described in detail below in conjunction with the accompanying drawings.

Fig. 1 is a topological diagram of the network composition of the data grid system of the present invention, which is composed of server group 1 and access client 2, wherein the server group is composed of metadata server 3 and multiple replica storage servers 4 , wherein the metadata server 3 is connected to the replica storage server 4. The access client 2 is connected to the metadata server 3 and the copy storage server 4. The read and write operations of the client are composed of two stages, the first is the metadata operation with the metadata server, and the second is the data exchange with the copy storage server. operate.

As shown in Figure 2, the metadata server consists of 6 modules, namely: replica view update module 10, replica view storage module 11, identity authentication module 12, server distance comparison module 13, replica view selection module 14, error handling Module 15.

The functions of each module are as follows:

Replica view update module 10: receives the replica update report from the replica storage server, finds and deletes outdated replica view information, creates new replica view information, modifies the data segment description information of the replica view, and sends an update confirmation to the replica storage server.

Copy view storage module 11: store copy view information, and provide view query service for other modules. The complete view information describing a replica consists of the address of the replica storage server and the data segment description information of the available replicas stored on the replica storage server. The data segment description information consists of the starting offset and the data segment The length indicates that the address of the duplicate storage server is indicated by the domain name or IP address of the duplicate storage server.

Identity authentication module 12: perform identity authentication processing based on the file handle for the user's read-write view request, that is, judge whether the file handle provided in the request is the file handle of a legitimate user.

Server distance comparison module 13: Estimate the logical distance between each replica storage server and the user who initiates the read/write request. The longer the distance, the smaller the available bandwidth of the communication link between the server and the user, and vice versa.

Copy view selection module 14: select the copy view with the smallest distance from the user according to the user's read and write request and the output result of the server distance comparison module; when the user performs a file read operation, the copy view with the smallest distance may It is composed of multiple data segments on multiple replica storage servers; when a user performs a file write operation, the replica view with the smallest distance is a data segment on the replica storage server closest to the user.

Error handling module 15: when identity authentication fails, record an exception, and send a request failure response to the client.

As shown in FIG. 3 , the access client is composed of five parts: a user access interface module 200 , a view request module 201 , a view analysis module 202 , a read-write execution module 203 and an error handling module 204 .

The functions of each module are as follows:

User access interface module 200: receiving the file read and write request sent by the user, and returning the execution result to the user.

The view request module 201: sends a read and write view request to the metadata server according to the output parameters of the user access interface module, and obtains copy view information required for specific read and write operations.

Replica view parsing module 202: analyze the replica view information returned by the metadata server, group the replica view information according to the servers where the data segments in the replica view information are located, and the data segments stored on the same server belong to the same group, the same Each data segment of the group is sorted in the group according to the starting offset; for the read operation, the data of each data segment is obtained from the relevant copy storage server through the read and write execution module 203, and a complete data segment is formed and returned to the user access interface module 201 ; For the write operation, the user data is written into the corresponding data segment of the corresponding replica storage server through the read-write execution module 203 ;

Read and write execution module 203: Interact with the replica storage server to execute read and write file operations.

Error handling module 204: process errors generated when reading and writing view requests and executing file reading and writing.

Figure 4 shows the flow chart of the user's file writing operation. In the access client, the user access interface 200 simply encapsulates the user's file writing request in step 300, and the view request module 201 sends the request to the element in step 310. The data server requests the view information required for the write operation; in the metadata server, the identity authentication module authenticates the user’s view request in step 320, and if the authentication is passed in step 330, then go to step 340 and the server distance The comparison module 13 estimates and selects the replica storage server closest to the user, and proceeds to step 350, where the replica view selection module 14 generates view information and sends it to the access client. If the authentication is not passed in step 330, then proceed to step In 360, the error processing module 15 sends an error report to the access client; in the access client, in step 370, the view parsing module 202 receives the view response from the metadata service, and in step 380, if the view is invalid view information, then Go to step 440 and carry out error handling by the error handling module 204, and go to step 450 to prompt the user to write failure by the user access interface 200, if in step 380 it is judged that the view is a valid view, then go to step 390 by the write execution module 203 initiates a write execution request to the copy storage server; in the copy storage server, step 400 checks the validity of the write execution request from the access client, if it is an invalid request, then return to the access client error message and go to step 420, if it is valid request, then perform a write operation, send a copy update report to the metadata server, and update the copy view by the copy view update module 10 of the metadata server in step 410, and send it to the access client in step 400 after the update is completed Write the execution result; in the access client, in step 420, the write execution module 203 receives the write execution result of the copy storage server, and in step 430, it is judged whether the write execution is successful, and if it is successful, then go to the user access interface 200 in step 450 Return a user writing response to the user. If it is judged that the writing execution fails in step 430, then go to step 440, and the error handling module 204 performs error handling, and go to step 450, and the user access interface 200 prompts the user to write failure.

Fig. 5 is the flowchart of the user's file reading operation. In the access client, the user access interface 200 simply encapsulates the user's file reading request in step 500, and then transfers to the metadata server by the view request module 201 in step 510. Request the view information required for the read operation; in the metadata server, in step 520, the identity authentication module performs identity authentication on the user's view request. 13 Estimate the distance between each replica storage server and the user, and go to step 550, and the replica view selection module 14 generates view information according to the distance information estimated in step 540 and sends it to the access client. If the authentication is not passed in step 530, Then turn to step 560 and send the error report to the access client by the error processing module 15; in the access client, the view analysis module 202 receives the view response from the metadata service in step 570, and if the view is invalid in step 580 view information, then go to step 620 to handle the error by the error handling module 204, and go to step 650, and the user access interface 200 prompts the user to write failure, if it is judged that the view is a valid view in step 580, then go to step 590 The read execution module 203 initiates a read data request to the copy storage server involved in the view; in the copy storage server, step 600 checks the validity of the read execution request from the access client, and returns it to the access client if it is an invalid request The error information goes to step 610, if it is a valid request, then execute the read data operation, and send the read data result to the access client; in the access client, the read execution module 203 receives the read data result of the replica storage server in step 610 , in step 610, it is judged whether the read data is successful, and if it is successful, then go to step 640, and the view parsing module stitches the read data according to the view information; Return the user to read the response to the user, if it is judged that the read data fails in step 610, then go to step 620 to perform error handling by the error handling module 204, and go to step 650 to prompt the user to read the failure by the user access interface 200.

Claims (2)

1. a method that is used for the operating writing-file of multi-copy data grid system is characterized in that, implementation step is as follows successively:

Step (1). the user capture interface module of the copy access client of user in described system is submitted the written document request to, and this request comprises: the sign of institute's written document, the data that will write and the initial address side-play amount of institute's write data section and the length of data segment;

Step (2). after described user capture interface module is carried out simplified package to described written document request, be sent to the view request module of described copy access client, this view request module authentication module in the meta data server in described system is sent and is write view request;

Step (3). the authentication module of described meta data server is carried out simple authentication to the described view request of writing, if: not by authentication, then error message is issued error handling processing module in this meta data server, behind recording exceptional, the view parsing module in described system in the copy access client sends writes view request inefficacy response; If: by authentication, then correct information of authentication and user's the view request of writing is sent to server in this meta data server apart from comparison module, a copy storage server of the user's who initiates to write view request logical reach minimum is estimated and selected and leave to this server according to the domain name or the IP address of each copy storage server of storage in advance, apart from comparison module;

Step (4). the copy view selection module in the described meta data server is according to writing view request from described server apart from copy storage server and the user that comparison module obtains, in the copy view that copy view memory module from described meta data server is provided, choose the copy view, generate corresponding copy view information, and the result is returned to the view parsing module in the copy access client in the described system;

Step (5). the view parsing module of described copy access client receives that copy view information that the copy view selection module from described meta data server is returned or error handling processing module return writes view request and lost efficacy after the response, and whether check is effective view information; If the error handling processing module of this copy access client then issued error message by invalid view, and returns error message by described user capture interface module to the user; If effective view, then resolve described copy view information, according to the data segment place server in the copy view information described copy view information is divided into groups, make the data segment that is stored on the same server belong to same group, again each data segment of same group is organized internal sort by the size of the side-play amount of initial address, and the read-write Executive Module that group view information is sent in the described copy access client carries out concrete data writing operation;

Step (6). the read-write Executive Module of described copy access client sends write data requests according to the group view information of receiving from described view parsing module to the copy storage server;

Step (7). the validity of the write data requests that the read-write Executive Module of the described copy storage server check of step (6) from described copy access client received; If the invalid data request, then the read-write Executive Module to described copy access client returns error message; If effectively request is then carried out data writing operation to the data copy, after writing successfully, the copy view update module of described copy storage server in described meta data server sends the copy updating record;

Step (8). after the copy view update module of the described meta data server of step (7) is received the copy updating record that obtains from the copy storage server, by the copy view memory module in the described meta data server copy view information of correspondence is upgraded operation, so that provide copy view accurately to described copy view selection module, if copy view update failure, then described copy view update module is returned copy to the described copy storage server of step (7) and is upgraded the failure confirmation, and then described copy storage server is rolled back to the state before the write operation, and sends the write data failure response to the read-write Executive Module of the described copy access client of step (6);

Step (9). if the described copy view update success of step (8), then the described copy view update of step (8) module is returned copy to the described copy storage server of step (7) and is upgraded successful confirmation, and this copy storage server sends the write data success response to the read-write Executive Module of the described copy access client of step (6);

Step (10). the read-write Executive Module of the described copy access client of step (6) is received after the write data response of the described copy storage server of step (7), response is judged to write data, if write data failure response, then error message is issued the error handling processing module of this copy access client, and returned error message to the user by described user capture interface module; If the write data success response is then returned the write data successful information to described view parsing module, this view parsing module returns the written document response by described user capture interface module to the user.

2. the method that is used for the operating writing-file of multi-copy data grid system according to claim 1, it is characterized in that, when described method is used for read operation, after authentication is passed through, by server apart from comparison module according to the logical reach Information Selection between copy storage server and the user and generate corresponding view information.

Images