WO2012083797A1

WO2012083797A1 - Write lock method and system for distributed file system

Info

Publication number: WO2012083797A1
Application number: PCT/CN2011/083694
Authority: WO
Inventors: 朱鹏; 江凌生; 王瑜
Original assignee: 中兴通讯股份有限公司
Priority date: 2010-12-20
Filing date: 2011-12-08
Publication date: 2012-06-28
Also published as: CN102088484B; CN102088484A

Abstract

A write lock method and system for a distributed file system are disclosed. The method comprises the following steps: when an upper-layer data location register server receives a request from a file request client for writing an opened file, the upper-layer data location register server extracting data block path information of the file, and returning the data block path information to the file request client; a lower-layer lock server, according to a request sent from the file request client for locking a specific data part in a specific data block of the file, returning write lock permission information of the specific data part of the specific data block to the file request client. The present invention not only guarantees improvement of the throughput capacity of lock submission in the distributed file system without increasing any hardware investment, but also has good extensibility as synchronous capacity expansion can be achieved when the capacity expansion of the distributed file system is performed.

Description

Write lock method and system for distributed file system

The present invention relates to the field of data storage, and in particular to a write locking method and system for a distributed file system. Background technique

In large distributed file systems, in order to achieve massive concurrency and ensure the consistency of file system data, a unified locking node is usually selected to centrally lock and manage the data stored in the distributed file system. The locking node can be a single server or cluster. For convenience of description, the locking node (locking server or cluster) is simply referred to as a locking server. However, with the rapid development of the multimedia industry, due to cost, reliability and other considerations, for large distributed file systems, more and more vendors or enterprises use the cloud to store. Cloud storage refers to a system that integrates a large number of different types of storage devices on a network through application software through cluster applications or distributed file systems to provide data storage and service access functions. Cloud storage is a cloud computing system with data storage and management at its core. In cloud storage, both security and cost are considered, so how to deploy the cloud is key. Especially nowadays, the scale of the cloud is getting bigger and bigger, more and more clouds will be distributed in multiple places, and how to deploy the distributed file system locking server will become very complicated.

In cloud storage, when a single locking server is used to lock the core part of the distributed file system, when multiple programs access different file offsets of different files in the distributed system at the same time, the locking server needs to be accessed sequentially to obtain the pair. Write lock permission for a specific file offset for a particular file. At this point, the access throughput of the distributed file system is greatly limited, resulting in a drop in throughput performance. And when multiple programs access the same file offset of the same file in the distributed system at the same time The first accessed program gets the write lock permission for the specific file offset for that particular file. Then, at this time, other access programs can only wait in the cache. Only when the first accessed program releases the write lock permission for the specific file offset of the particular file, the other program can obtain the write lock permission for the specific file offset of the specific file. Therefore, a single lock server becomes a bottleneck for giant cloud writes. Summary of the invention

The present invention is directed to the deficiencies of the prior art, and provides a write locking method and system for a distributed file system, which not only ensures that the throughput of the distributed file system lock submission is improved without increasing the hardware investment, but also can With the expansion of the distributed file system and simultaneous expansion, it has good scalability.

In order to solve the above technical problems, the present invention provides the following technical solutions:

A write locking method for a distributed file system includes the following steps:

When the upper layer data location registration server receives the request for writing the open file sent by the file requesting client, the upper layer data location registration server extracts the data block path information of the file, and returns the file request client to the file;

The lower layer locking server returns, according to the file request, a lock request of the specific data part of the specific data block of the specific data block according to the data block path information, and the write lock permission information of the specific data part of the specific data block is returned. Request the client for the file.

In an embodiment of the invention, the method further includes:

When the upper layer data location registration server extracts the data block path information of the file and finds that there is no data block path information of the file, the upper layer data location registration server creates a unique complete data block path for the file, and records Data block path information of the file, and distributes write lock rights of different data blocks on the data block path of the file to the lower layer lock server.

In an embodiment of the invention, the method further includes: The file requesting client sends a lock request to the lower layer lock server for a specific data portion of the specific data block;

Before the file requesting client sends a lock request to the lower-level locking server for the specific data portion of the specific data block, the method further includes:

The file requesting client determines a lock condition of the specific data portion in its own local cache, and when the specific data portion is already in its own local cache and has been locked, the file request client does not The lower layer lock server sends a lock request; when the specific data portion is not locked, the file request client sends a lock request to the lower layer lock server.

In an embodiment of the present invention, before the file requesting, by the lower layer locking server, the lock request of the specific data part of the specific data block sent by the client, the method further includes: creating a data by the lower layer locking server Block lock status information table;

The lower lock server creates a data block lock status information table as:

When the lower layer locking server receives the lock request sent by the file requesting client to the specific data portion of the specific data block, the lower layer locking server determines whether it is currently receiving the specific data for the first time. Block lock request, when it is the first time, adding the specific data block and its corresponding lock request information to the data block lock status information table; when not for the first time, locking the data block The lock information of the specific data portion of the specific data block is retrieved in the status information table.

In an embodiment of the present invention, after the lower-level locking server returns the write-locking right information of the specific data portion of the specific data block to the file requesting client, the method further includes:

Determining, by the lower-layer locking server, whether the lower-layer data block server is notified according to the received lock request request of the file request client and the write lock permission information of the specific data portion of the specific data block returned to the file request client File request client needs Sending a specific data portion to the file requesting client;

The determining whether to notify the lower layer data block server to send the specific data portion required by the file requesting client to the file request client is:

After the lower layer locking server receives the read/write lock request sent by the file requesting client to the specific data portion of the specific data block, the lower layer locking server checks the specific content in the current data block lock status information table. a locking condition of the data portion, when the specific data portion is not locked, marking the locking state of the specific data portion as locking in the data block locking state information table, and notifying the lower layer data block server to return the specific The data portion requests the client for the file; when the specific data portion has been locked, performing a lock failure policy;

After the lower layer lock server receives the write lock request sent by the file request client for the specific data portion of the specific data block, the lower layer lock server checks the lock status of the current specific data portion, when The specific data portion is not locked, the lock status of the specific data portion is marked as locked in the data block status information table, and the lock is successfully returned; when the specific data portion has been locked, the lock is performed Failure strategy.

In an embodiment of the present invention, the file requesting client sends a read/write lock request or a write lock request to the lower layer lock server for the specific data portion of the specific data block as: the file request client check a hit situation in the local cache of the specific data portion, when the specific data portion is not in the local cache, determining that the specific data portion is not hit, and sending a read/write lock request to the lower layer lock server; When the specific data block is in the local cache and is not locked, it is determined that the specific data portion is hit, and a write lock request is sent to the lower layer lock server.

The present invention also provides a write lock system for a distributed file system, including a file request client, a lower layer data storage module, and an upper layer data location registration selection path module and a lower layer data locking module disposed in the system; The upper layer data location registration selection path module is configured to receive a request for writing an open file sent by the file request client, and extract the data block path information of the file and return the file request request to the client;

The lower layer data locking module is configured to: according to the file requesting a lock request sent by a client to a specific data portion of a specific data block in the file, returning, to the write permission authority information of the specific data portion of the specific data block The file requests the client.

In an embodiment of the present invention, the upper layer data location registration selection path module is further configured to: when extracting data block path information of the file and discovering that there is no data block path information of the file, Creating a unique complete data block path, recording data block path information of the file, and distributing write lock rights of different data blocks on the data block path of the file to the lower data lock module.

In an embodiment of the present invention, the file requesting client is configured to determine that the specific data part is in itself before sending a lock request to the lower layer data locking module for the specific data part of the specific data block. a lock condition in the local cache, when the specific data portion is already in its own local cache and has been locked, then no lock request is sent to the underlying data lock module; when the specific data portion is not locked, A lock request is sent to the underlying data locking module.

In an embodiment of the present invention, the lower layer data locking module is specifically configured to: according to the file request client, a lock request for a specific data portion of the specific data block, to create the data block lock status. The information table, when receiving the file requesting the client to send a lock request for the specific data portion of the specific data block, determining whether it is currently receiving the lock request for the specific data block for the first time, when The first time, the specific data block and its corresponding lock request information are added in the data block lock status information table; when not the first time, the data block lock status information table is retrieved Lock request information for a specific data portion.

In an embodiment of the present invention, the lower layer data locking module is further configured to receive according to And the file requesting the client's lock request and the write lock permission information of the specific data portion of the specific data block returned to the file request client, determining whether to notify the lower layer data storage module to request the client from the file The specific data portion required is sent to the file request client;

The lower layer data locking module is configured to: after receiving the read/write lock request for the specific data portion of the specific data block sent by the file requesting client, check the specific content in the current data block locking state information table. a locking condition of the data portion, when the specific data portion is not locked, marking the locking state of the specific data portion as locking in the data block locking state information table, and notifying the lower layer data storage module to return the The specific data portion requests the client for the file; when the specific data portion has been locked, executing a lock failure policy; when receiving the file requesting the client to send the specific data portion of the specific data block After the write lock request, checking a lock condition of the specific data portion in the current data block lock status information table, and if the specific data portion is not locked, the specific content is included in the data block status information table The lock status of the data portion is marked as locked, and the lock is returned successfully; when the specific data block has been Set, perform the lock failed strategy.

In an embodiment of the present invention, the file requesting client is specifically configured to check a hit situation of the specific data part in a local cache, and when there is no specific data part in the local cache, determining that there is no Hiting the specific data portion, and sending a read/write lock request to the lower data lock module; when the specific data portion is in the local cache and is not locked, determining to hit the specific data portion, and The lower data locking module sends a write lock request.

Because the above technical solution is adopted, the present invention has the beneficial effects of: determining, by the upper layer data location registration server, a unique data block path in one or more files requested by the client for the file, thereby data of the file The write lock permission of different data blocks on the block path is distributed to the lower lock server, and the data block path information of the file is returned to the a file requesting client; the file requesting client initiates, according to the data block path information obtained from the upper data location registration server, a request message for writing a specific data portion of a specific data block in the file to be written to the lower layer locking server, And obtaining write lock permission information of the specific data block from the corresponding lower layer lock server. In this way, the upper data storage location storage server distributes the write lock permission of all the data blocks of the file to one or more lower-level locking servers in the lower layer, and discards the traditional lock of the distributed file system by using a single locking server. The method, on the one hand, reduces the load of the locking server of the upper layer data location registration server; on the other hand, it also realizes the write lock permission of the data stored in the corresponding lower layer data block server in the lower layer locking server management distributed system, so that Increase the throughput of distributed file system lock submissions when adding distributed file system hardware resources. Meanwhile, the locking method provided by the present invention can be used in a distributed file system of one or more upper layer data location registration servers, and therefore, the locking method can improve the system throughput capability along with the distributed file system. Capacity expansion and simultaneous expansion, with good scalability. DRAWINGS

1 is a diagram showing the working architecture of an embodiment of the present invention;

2 is a flow chart of the operation of another embodiment of the present invention. detailed description

The present invention will be further described in detail below with reference to the accompanying drawings.

At present, in order to solve the bottleneck problem of a giant cloud write caused by a single lock server, as shown in FIG. 1, the present invention provides a write lock method and system for a distributed file system, the main idea of which is: in a distributed file system The lower layer uses multiple decentralized locking servers to collectively lock data stored in the distributed file system. Compared with the conventional single lock server, the present invention distributes the lock permission of the data stored in the distributed file system to be locked to a plurality of different file system data servers in the lower layer. Specifically, the file system data server mainly includes The lower layer locking server and the lower layer data block server, wherein the lower layer data block server is mainly used to store the data block information in the distributed file system, and the lower layer locking server is mainly used to lock the data stored in the lower layer data block server, that is, The lock permission of the data in the data block server is managed. The lock permission mentioned here mainly refers to the write lock permission of data in the distributed file system. Specifically, the write locking method determines a unique complete data block path through a file of one or more files requested by the upper layer data location registration server for the file request client to open, so that different data blocks on the data block path of the file The write lock permissions are distributed to multiple lower-level locking servers. Actually, in the upper layer data location storage server, a large amount of data block path information stored in the lower layer data block server is stored, and within the system, the data block path information stored in the lower layer data block server is also referred to as metadata. . At the upper data location registration server, a dedicated locking server is used to lock the metadata. In the lower layer, a plurality of distributed lower-level locking servers are used to lock the data blocks stored in the lower-level data block server. When the lower layer locking server receives the lock request from the file requesting client for the specific data portion of the specific data block in the file, the lower layer locking server returns the write lock permission information of the specific data portion of the specific data block to the file request. Client.

The write locking method provided by the present invention uses a plurality of distributed lower-level locking servers to lock data in the distributed file system, mainly to reduce the load of the locking server of the upper-layer data location registration server. At the same time, the upper layer data location registration server distributes the write lock permission of different data blocks in the file requested by the client requesting client to the lower lock server, so that the lower file system data server is no longer a buffer of the upper lock server. Realize the write lock permission of the underlying lock server to manage data in the distributed file system. The write lock method can be widely used in a single or multiple upper layer data location registration server. If used in a plurality of upper layer data location registration servers, each upper layer data location registration server does not have a primary standby relationship.

In an embodiment of the present invention, when the upper layer data location registration server extracts a certain When the data block path information of the specific file is found and the data block path information of the file is not found, for example, the data block path information of the file A is not found, it indicates that the upper layer data location registration server is the first time to receive the file request client. The write open request for the specific file A is sent, then the upper data location registration server first creates a unique complete data block path for the specific file A, and records the data block path information of the file A. Through the creation of the file data block path, the write lock permission of different data blocks on the data block path of file A is distributed to the lower layer lock server, thereby achieving the finest granularity (byte level) of locking. The data block path can describe path information of all data blocks in the file A, and mainly includes which data blocks are located in the file data block path and in which lower layer data block server the data blocks are stored. It should be noted that at this time, the lower-level locking server does not know that it manages the write lock permission of these data blocks.

When there is a file requesting the client to send a request to write the open file A to the upper layer data location registration server, since the only complete data block path of the file A has been created at this time, the upper layer data location registration server can extract the file A. The block path information is returned to the file request client. It should be noted that the file requesting client that sends the write open file A request for the first time and the file request client that sends the write open file A request for the second time can request the client for the same file, or can request for different files. Client. At the same time, when the two files request the client to write and lock the same data block in the file A, they will access the same underlying data block server that stores the data block.

In an embodiment of the present invention, the file requesting client determines, before the lower layer locking server sends a lock request for the specific data portion of the specific data block in the file A, the specific data portion is in the file request client local cache. Lock the situation. Here, for convenience of description, a specific data block is simply referred to as a data block B. If the data block B specific data portion is already in the file request client local cache and has been locked by the file request client, the file request client does not need to send a lock request to the lower layer lock server, and can directly start writing the data block B specific Operation of the data part; if the specific data part of the data block B is not locked, there may be two cases at this time, one is The file request does not have a specific data portion of the data block B in the local cache of the client, and in another case, there is a specific data portion of the data block B in the local cache of the file requesting client, but the specific data portion is not locked, for this In either case, the file requests the client to send a lock request to the lower level lock server.

In an embodiment of the present invention, the lower layer locking server creates a data block lock state information table according to a lock request sent by the client to the specific data portion of the specific data block, in a specific manner: when the lower layer locks the server C Upon receiving the file requesting client to send a lock request for the specific data portion of the data block B, the lower layer locking server C needs to determine whether it is currently the first time the file request client receives the lock on the data block B. The request, if it is, the lower level lock server C knows that it manages the write lock permission of the data block B. It also indicates that there is no information about the data block B in the existing data block lock status information table stored in the lower layer lock server C. At this time, the lower layer locking server C adds the entry of the data block B and the corresponding lock request information of the data block B in the data block lock state information table; if not, the lower layer lock server C only needs to lock the state information in the current data block. The table can retrieve the lock request information of the specific data portion of the data block B.

In an embodiment of the present invention, it is assumed that the lower-layer locking server C does not receive the lock request for the specific data portion of the data block B sent by the file requesting client for the first time, that is, at this time, the lower-layer locking server C Retrieving the relevant write lock permission of the specific data portion of the data block B according to the received file requesting the different lock request of the client, and returning the write lock permission information of the specific data portion of the client data block B to the file request, and determining whether to notify the lower layer The lower data block server corresponding to the lock server C sends the specific data portion required by the file request client to the file request client, specifically: when the lower layer lock server receives the file request client and sends the data block B specific When the data portion D is read or written by the lock request, the specific data portion D is the offset of the file A for the user. The lower lock server checks the lock status of the specific data portion D in the current data block lock information table. If the specific data portion D is not locked, the table is If the specific data portion D has not been locked by another program, the lock status of the specific data portion D is marked as locked in the data block lock status information table, and the lower data block server is notified to return the specific data portion D to the file request client; It should be noted that for the same data block, it will contain a lot of data information, and the file request client needs to lock only a few pages of the data information. Here, the data information is divided into a plurality of parts, and then the file requesting client needs to lock one of the plurality of parts. If two files request the client to write different data parts in the same data block B at the same time, then the two files request the client can write at the same time. In other words, from the user's point of view, if two users need to write locks to different file offsets in the same file, then the two users can simultaneously write. If the specific data portion D has been locked, indicating that the specific data portion D has been locked by another program, the corresponding lock failure strategy is used;

When the lower layer locking server receives the write lock request for the specific data portion of the data block B sent by the file requesting client, the lower layer locking server checks the locking condition of the specific data portion in the current data block locking information table, if the specific data portion is not Locking, in the data block state information table, marking the locking state of the specific data portion as locking, and returning the file requesting client to the successful locking; if the specific data portion has been locked, using the corresponding locking failure policy, specifically Say, the lock failure policy includes lock failure or other strategies.

Further, when the file requesting client receives the message that the specific data portion of the data block B is successfully locked, the message that the write lock is successful is returned to the operation user, and the user can perform the operation of starting writing or modifying the specific data portion of the data block B. It is. If the file requesting client receives a lock failure policy corresponding to the execution of the specific data portion of the data block B, the file request client may choose to wait for the lock to continue, or directly return the write lock failure to the operation user, at this time, the user does not have Permission to write to a specific data portion of block B.

In fact, the file request client checks the specific data portion locally before sending the read/write lock request or the write lock request to the specific data portion of the data block B to the lower layer data storage server C. Hits in the cache. If there is no specific data part in the local cache, that is, there is no data stored in the local cache requesting the client to lock the data, it is judged that the specific data block part is not hit, and the read/write lock request is sent to the lower layer lock server C; There is a specific data part, that is, the file requesting client performs a read operation or a write operation on the specific data portion of the data block B, but only the specific data portion has not been locked by the file request client at this time, and the J J judges the hit. A specific data portion, and a write lock request is sent to the lower layer lock server C. That is to say, if a file client has previously performed a read operation or a write operation on a specific data block portion, when it requests the specific data portion again, it only needs to send a write lock request to the lower layer lock server.

As shown in FIG. 2, it is a flowchart of a method according to an embodiment of the present invention. Referring to FIG. 2, an embodiment of the present invention is explained in detail. For convenience of description, the file request client needs to lock the specific data portion of the data block B in the file A as an example, wherein Fac represents the file request client, and Fas represents the lower file system data server (including the lower layer data block server and The lower layer locks the server), and Fir represents the upper layer data location registration server. The specific steps are as follows:

SI: Fac request information written to Fir to open file A;

S2: Fir receives the request information of the write open file A issued by the Fac and extracts the data block path information of the file A, that is, the chunk information. If Fir cannot extract the chunk information of the file A, Fir determines whether it is the first time to receive the Fac. The incoming request to open file A is written, and a unique complete chunk that can completely describe all the data blocks in file A is created for file A, and the chunk information is recorded. If Fir can extract the chunk information of file A, Fir directly extracts the chunk information of file A and returns it to Fac;

S3: Fac determines the locking condition of the specific data part of the data block B in the local cache. If the specific data block part is in the local cache of the Fac and has been locked, go to step S4; if the specific data part is not locked, go to Step S5;

S4: It has been determined that the specific data portion has been locked, and the Fac returns the user to successfully lock, and can enter The operation of writing a specific data portion of the data block B;

S5: Fac judges the hit of the specific data part of the data block B in the local cache, if it is not, then goes to S6, if it hits, it goes to S9;

S6: Fac sends a request for reading and writing a block data block B to a lower data storage server in the Fas;

S7: Fas checks whether the specific data portion of the current data block lock status information table data block B is locked. If not locked, the lower lock server marks the lock status of the specific data portion of the data block B in the data block lock status information table. To lock and return the specific data portion to the file request client; if it is locked, execute the lock failure policy;

S8: Fac checks the status returned by Fas. If the specific data part is returned, the operation returns to the user that the lock is successful and starts writing. If the execution lock failure policy is received, the user may return the lock failure or wait for a while to continue. Request a lock.

S9: Fac sends a request to the lower data storage server in the Fas to write a lock data block B specific data portion;

S10: Fas checks whether the specific data portion of the current data block lock status information table data block B is locked. If not locked, the lower lock server marks the lock status of the specific data portion of the data block B in the data block lock status information table. To lock and return the lock to the file request client successfully; if it is locked, execute the lock failure policy;

Sl l: Fac checks the status returned by Fas. If the return lock is successful, it returns the operation that the lock is successful and starts writing. If the lock execution failure policy is received, the lock failure may be returned to the user or wait for a period of time to continue to request the lock. .

In the meantime, the present invention also provides a write lock system for a distributed file system, comprising an upper layer data location registration selection path module, a lower layer data locking module and a lower layer data storage module disposed in the system;

The upper layer data location registration selection path module is configured to receive the file request client to send the write Opening a request for a file, and extracting data block path information of the file is returned to the file request client;

a lower layer data locking module, configured to: according to the file requesting a lock request sent by a client to a specific data portion of a specific data block in the file, returning write lock authority information of the specific data portion of the specific data block to the File request client.

It should be noted that, in the foregoing embodiment, for convenience of description, the file requesting client needs to lock the specific data portion of the data block B as an example for description. In fact, in a distributed file system, a large number of file system data servers store a large amount of data stored in the form of data blocks. The upper data location registration server divides the same file into multiple data blocks and stores them in the same or different lower file system data servers, usually in different file system data servers, specifically in the data block server. Inside. That is to say, there are many data blocks from different files stored in the same underlying file system data server. Once the upper data location registration server determines a unique complete data block path for a file, all file request clients need to lock a particular data portion of a data block in the file. This unique block path is accessed and the corresponding lock operation is performed. That is to say, at this time, the lock permission of the data block has been distributed to the underlying file system data server. When the upper layer data location registration server receives the file request client's write request to open a file lock request, only the determined one copy data block path is selected, so that the load of the upper data location registration server is not raised by itself. Since then, write locks on a block of data in a distributed file system have nothing to do with the upper data location hosting server.

Since there are many data servers in the lower file system, the data blocks managed by each file system data server are limited. Therefore, the write lock permission of the data blocks that each file system data server needs to manage is limited, so the lower file system is limited. The load on the data server will not increase too much, so it will not affect the normal operation of the underlying file system data server. Achieve a single lock server in an existing distributed file system by using multiple file system data servers Locking the load improves the throughput of the distributed file system lock submission without increasing the hardware investment of the distributed file system itself. In practical applications, you can select a data location registration server (single master) or multiple data location registration servers (multi-master). If a plurality of data location registration servers are used in the upper layer, and data write lock permissions in the distributed file system are distributed to a large number of lower file system data servers through a plurality of data location registration servers, then in the distributed file system The growth of the underlying file system data server, the locking system can be expanded with the expansion of the distributed file system, and thus has good scalability.

On the premise of writing and opening the same file, except for the first time that the upper layer data location registration server sends a file request client that writes the file request for opening the file, each of the other file request clients only needs to register with the upper layer data location. Once the server communicates once, the data block path information of the file where the data portion to be locked is located can be obtained. If the file requests the client to obtain the write lock permission of the data block again, the file request client only needs to send a lock request to the lower layer data storage server, and the write lock permission of the data block can be obtained from the lower lock server. This reduces the number of times the file request client communicates with the upper data location hosting server. The write lock permission of the data block includes write lock permission of all data stored in the data block. For the same file request client, only one communication with the upper data location register is required, so that the upper data location registration server can basically maintain the same performance as the file system that does not support the write lock, thereby reducing the performance. Unnecessary communication overhead. It is to be understood that the specific embodiments of the invention are limited only by the description. It is to be understood by those skilled in the art that the present invention may be practiced without departing from the spirit and scope of the invention.

Claims

Claim

A write lock method for a distributed file system, wherein the method comprises: when an upper layer data location registration server receives a request for writing an open file sent by a file request client, the upper layer data location registration server extracts Data block path information of the file, and returned to the file request client;

And the lower layer locking server requests, according to the file, a lock request information of a specific data part of the specific data block in the file according to the data block path information, and writes the lock permission information of the specific data part of the specific data block. Return to the file request client.

2. The method according to claim 1, wherein the method further comprises: when the upper layer data location registration server extracts data block path information of the file and finds that there is no data block path information of the file The upper data location registration server creates a unique complete data block path for the file, records data block path information of the file, and distributes write lock permissions of different data blocks on the data block path of the file to The lower layer is locked in the server.

The method according to claim 1, wherein the method further comprises: the file requesting client transmitting, to the lower layer locking server, a specific data portion of the specific data block according to the data block path information Lock request

Before the file requesting client sends a lock request for the specific data portion of the specific data block to the lower layer lock server according to the data block path information, the method further includes: the file requesting client determining the The locking situation of a particular data portion in its own local cache, when the particular data portion is already in its own local cache and has been locked, then the file requesting client does not send a lock request to the underlying lock server.

The method of claim 3, wherein the method further includes: before the file requesting, by the lower-layer locking server, a lock request from the client to the specific data portion of the specific data block, the method further includes: The lower lock server creates a block lock status information table; The lower lock server creates a data block lock status information table as:

The method of claim 4, after the lower-level locking server returns the write-locking right information of the specific data portion of the specific data block to the file requesting client, the method further includes:

Determining, by the lower-layer locking server, whether to notify the lower-layer data block server according to the received lock request request of the file request client and the write lock authority information returned to the specific data portion of the specific data block of the file request client The specific data portion required by the file requesting client is sent to the file requesting client;

After the lower layer locking server receives the write lock request sent by the file requesting client to the specific data portion of the specific data block, the lower layer locking server checks the current location a locking condition of the specific data portion in the data block locking state information table, and when the specific data portion is not locked, marking a locking state of the specific data portion as locking in the data block state information table, And returning the lock success; when the specific data portion has been locked, the lock failure policy is executed.

The method according to claim 5, wherein the file requesting client sends a read/write lock request or a write lock request to the lower layer lock server to the specific data portion of the specific data block as:

The file requesting client checks a hit condition of the specific data portion in the local cache, and when there is no specific data portion in the local cache, determining that the specific data portion is not hit, and locking the server to the lower layer Sending a read/write lock request; when there is the specific data portion in the local cache and is not locked, determining to hit the specific data portion and transmitting a write lock request to the lower lock server.

A write lock system for a distributed file system, comprising a file request client and a lower layer data storage module, wherein the system further comprises an upper layer data location registration selection path module and a lower layer data lock module disposed in the system. ; among them,

The upper layer data location registration selection path module is configured to receive a request for writing an open file sent by the file requesting client, and extract the data block path information of the file and return the data to the file request client;

8. The system according to claim 7, wherein the upper layer data location registration selection path module is further configured to: when extracting data block path information of the file and discovering that there is no data block path information of the file Creating a unique complete data block path for the file, recording data block path information of the file, and different data blocks on the data block path of the file Write lock permissions are distributed to the underlying data lock module.

The system of claim 7, wherein the file requesting client is configured to determine, before sending a lock request for the specific data portion of the specific data block to the lower layer data locking module, a lock condition of a particular data portion in its own local cache, when the particular data portion is already in its own local cache and has been locked, then no lock request is sent to the underlying data lock module; when the particular data portion is not When locked, a lock request is sent to the underlying data locking module.

The system of claim 9, wherein the lower layer data locking module is configured to: according to the file request client, a lock request for a specific data portion of the specific data block, a data block lock status information table, when receiving a lock request from the file requesting client for a specific data portion of the specific data block, determining whether it is currently receiving the specific data block for the first time Locking the request, when it is the first time, adding the specific data block and its corresponding lock request information to the data block lock state information table; when not the first time, locking the state information in the data block The table retrieves the lock request information for the specific data portion.

The system of claim 10, wherein the lower layer data locking module is further configured to: request a lock request of the client according to the received file, and return the specific data to the file request client. Determining whether to notify the lower layer data storage module to send the specific data portion required by the file requesting client to the file request client by: the write lock permission information of the block specific data portion:

The lower layer data locking module is configured to: after receiving the read/write lock request for the specific data portion of the specific data block sent by the file requesting client, check the specific content in the current data block locking state information table. a locking condition of the data portion, when the specific data portion is not locked, marking the locking state of the specific data portion as locking in the data block locking state information table, and notifying the lower layer data storage module to return the Specific data portion to the text Requesting a client; when the specific data portion has been locked, executing a lock failure policy; after receiving the write request request sent by the client requesting the specific data portion of the specific data block, checking a lock condition of the specific data portion in the current data block lock status information table, and when the specific data portion is not locked, marking a lock status of the specific data portion in the data block status information table as Locked, and returns the lock successfully; when the specific data block has been locked, the lock failure policy is executed.

The system of claim 11, wherein the file requesting client is specifically configured to check a hit situation of the specific data part in a local cache, when the specific data is not in the local cache. Partially, determining that the specific data portion is not hit, and sending a read/write lock request to the lower data lock module; determining that the specific data is hit when the specific data portion of the local cache is not locked And sending a write lock request to the underlying data locking module.