KR100988107B1 - Distributed file system and file consistency management method using replica state matching - Google Patents

Abstract

Disclosed are a file consistency management method using a distributed file system and replica state matching. A method of managing file consistency using replica state matching may include storing and maintaining a replica and first replica status data by a data server, receiving a file operation request including second replica status data from a client server by the data server; Determining, by the data server, whether the first replica status data and the second replica status data match, and when the first replica status data and the second replica status data match, the data server files. It may include performing an operation.

Distributed File System, Replicas, Replica State Matching, Owner, Metadata

Description

File consistency management using distributed file system and replica state matching {DISTRIBUTED FILE SYSTEM AND FILE CONSISTENCY MANAGEMENT METHOD USING REPLICA STATE MATCHING}

The present invention relates to a file consistency management method using a distributed file system and replica state matching. More specifically, a distributed file system and a replica that perform file operations by matching copy states managed by servers constituting the distributed file system. A file consistency management method using state matching.

Recently, a method of building a distributed file system using a plurality of low price servers to store a large amount of file data has been proposed. This distributed file system has excellent cost and performance benefits because many low cost servers distribute and process file data.

However, in the case of using a low-cost server, a failure occurs frequently in file data. In other words, using a low-cost server has the advantage of reducing the hardware cost for constructing a distributed file system, but at the same time, there is a disadvantage that the computational power of the system is degraded due to the high frequency of failure caused by the low-cost server.

To overcome this drawback, distributed file systems may use replicas of files. However, the distributed file system is composed of a plurality of servers, so that the replica state recognized by each of the servers for the same replica may be different. Due to this problem, when a distributed file system performs a file operation, the contents of the same replica may vary from server to server, which may cause a reduction in file consistency of the entire system.

Thus, there is a need for a method that can perform file operations while maintaining file consistency throughout the system.

The present invention provides a distributed file system and a replica state matching that can maintain file consistency throughout the system by performing a file operation when the replica state data stored by all components of the distributed file system are identically matched. Provides a file consistency management method.

The present invention provides a file coherency management using distributed file system and replica state matching, which reduces the overhead incurred when a data server fails by storing and maintaining a replica state managed by a data server and a server state of another data server. Provide a method.

In the file consistency management method using replica state matching according to an embodiment of the present invention, a data server stores and maintains replica and first replica state data, and the data server stores second replica state data from a client server. Receiving a file operation request including; the data server determining whether the first replica state data and the second replica state data match and the first replica state data and the second replica state If the data matches, the data server may include performing a file operation.

In the file consistency management method using replica status matching according to an embodiment of the present invention, a method in which a client server stores and maintains first replica status data, and the client server includes a file including the first replica status data in a data server The method may include transmitting an operation request and receiving a file operation result from the data server when the first replica state data and the second replica state data stored in the data server match.

Data server according to an embodiment of the present invention is a data storage unit for storing and maintaining the replica and the first replica status data, a file operation request receiver for receiving a file operation request including the second replica status data from the client server, the A data matching determination unit determining whether the first replica state data and the second replica state data match, and when the first replica state data and the second replica state data match, the data server performs a file operation. It may include a file operation execution unit to perform.

The client server according to an embodiment of the present invention is a data storage unit for storing and maintaining first replica status data, a file operation request transmitter for transmitting a file operation request including the first replica status data to a data server and the When the first replica state data and the second replica state data stored in the data server match, it may include a file operation result receiving unit for receiving a file operation result from the data server.

According to the present invention, if the replica state data stored by all the components constituting the distributed file system are identically matched, the file operation is performed so that the distributed file system and the replica state can maintain file consistency throughout the system. A file consistency management method using matching is provided.

According to the present invention, a file using a distributed file system and replica state matching can reduce the overhead incurred when a data server fails by storing and maintaining a replica state managed by the data server and a server state of another data server. Consistency management methods are provided.

Hereinafter, with reference to the contents described in the accompanying drawings will be described in detail an embodiment according to the present invention. However, the present invention is not limited to or limited by the embodiments. Like reference numerals in the drawings denote like elements. The file consistency management method using replica state matching according to an embodiment of the present invention may be performed by a distributed file system.

1 is a view showing the overall configuration of a distributed file system according to an embodiment of the present invention.

Referring to FIG. 1, a distributed file system according to an embodiment of the present invention may include a client server 101, a metadata server 102, and a plurality of data servers 103. The number of data servers according to an embodiment of the present invention is not limited.

In the distributed file system according to an embodiment of the present invention, the client server 101 may be plural. For example, the metadata server 102 and the plurality of data servers 103 may constitute one cluster. In this case, the distributed file system may include at least one cluster.

The data server 103 may store a replica for each owner. For example, the owner may be determined based on a directory of the file as a unit for grouping files stored in the data server 103. In other words, a replica by owner refers to a set of files grouped by owner. The owner is described in detail in FIG. 2.

The metadata server 102 may store the owner-specific metadata composed of storage location information of a replica for each owner. In this case, the storage location information of the replica for each owner means a list of the data server 103 in which the replica for each owner is stored.

The client server 101 may request metadata about a predetermined owner from the metadata server 102 and receive the metadata. At this time, the predetermined owner means the owner to which the file to be processed by the client server 101 belongs. Then, the client server 101 may identify the data server 103 which is a copy storage location of the predetermined owner through metadata.

In this case, each of the client server 101, the metadata server 102, and the plurality of data servers 103 may store replica state data for a specific replica. In this case, the metadata server 102 may periodically store and maintain the replica state data of the latest state by periodically reflecting the replica state managed by each of the plurality of data servers 103. At this time, the replica status data indicates whether the replica is in a normal state (Normal: N), a failure state (Failure: F), a recovery state (R), or a migration state (M).

Since the client server 101 is a server corresponding to the user level, it is difficult to immediately reflect the replica status data of the metadata server 102, so that the system-wide file consistency may not be guaranteed. In order to solve this problem, the distributed file system according to an embodiment of the present invention uses client state 101, metadata server 102, and a plurality of data servers through replica state matching during file operations. (103) By controlling the copy state data managed by each to be identical to each other, it is possible to maintain the file consistency of the entire system.

As shown in FIG. 1, the client server 101 may transmit a file operation request to the data server 103 in which a copy of the owner is stored. In this case, the method of requesting a file operation from the data server 103 by the client server 101 may vary according to the content of the file operation. For example, a file write operation or a file deletion operation corresponds to a file operation request that changes the contents of the replica. In this case, the client server 101 may transmit a file operation request for changing the contents of the replica to all data servers 103 storing the replica of the same owner in order to maintain file consistency throughout the distributed file system.

At this time, the client server 101 may include the copy state data managed by the client server 101 in the file operation request, and transmit the file operation request to the data server 103. For example, the data server 103 may determine whether to match the replica state data stored in the client server with the replica state data included in the client server 101. At this time, the data server 103 may determine whether to perform a file operation according to the matching result of the replica state data. If the replica state data matches, the data server 103 may perform a file operation. If the replica state data does not match, the data server 103 may reject the file operation request.

In this case, the client server 101 may update its replica status data by referring to the replica status data of the metadata server 102. The client server 101 may resend the file operation request including the updated replica status data to the data server 103. Then, the data server 103 may again determine whether the replica status data is matched using the replica status data included in the retransmitted file operation request.

Finally, the process of updating the copy state data and transmitting the file operation request using the updated copy state data may be repeated until the copy state data is matched. Replica state matching is described in more detail with reference to FIGS. 3 to 8.

2 is a diagram illustrating an example in which files are grouped in units of owners based on directories according to an embodiment of the present invention, and copies of owners are stored in a data server.

Referring to FIG. 2, a file directory 201 stored in a data server is shown. At this time, the file directory 201 is composed of dir1 which is a subdirectory. At this time, file 1, file 2, and file 3 are stored in the subdirectory dir1. According to an embodiment of the present invention, an owner may be determined based on a directory of files. In other words, the owner may correspond to a directory of files actually stored in the data server.

As a result, the files of the data server can be managed according to the owner unit. That is, the distributed file system according to an exemplary embodiment of the present invention may manage the files efficiently by managing each file directory (that is, owner unit) rather than managing each file. Since the file is managed by the owner, which is a directory unit, even if the number of small files increases exponentially, the amount of metadata processed by the distributed file system is relatively less increased.

In this case, referring to FIG. 2, the directory dir / dir1 of the file may be determined as owner 1. Also, an owner different from owner 1 may be determined for the dir directory. If there are directories dir / dir2 and dir / dir3 in the file directory 201, each of the directories may be determined to be owner 2 and owner 3 different from owner 1. That is, files stored in the data server may be grouped according to owner units.

According to an embodiment of the present invention, the distributed file system may create a replica for each owner. At this time, the number of replicas created for each owner is not limited. Referring to FIG. 2, when the directory dir / dir1 is determined to be owner 1, three replicas X 202-1, 202-2, and 202-3 for owner 1 are generated. At this time, the number of replicas created for each owner is not limited.

Then, the data server may store any one of the replicas created for each owner. For example, when a plurality of replicas X are created for owner 1 and a plurality of replicas Y are created for owner 2, each data server may store one replica X or one replica Y. As a result, the data server can manage multiple files by storing one copy.

As can be seen in FIG. 2, when three replicas X are created for owner 1, replica X 202-1 is stored in data server 1 103-1, and replica X 202-2 is a data server. 2 (103-2). In addition, replica X 202-3 for owner 1 may be stored in data server 3 103-3. At this time, the data server 1 103-1, the data server 2 103-2, or the data server 3 103-3 may also store a replica of the owner 1 and another owner. That is, the number of replicas per owner stored in each data server is not limited.

Thus, in order to effectively handle frequent failures, the distributed file system according to an embodiment of the present invention may physically store a plurality of copies of one owner in another data server 103. As a result, even if a specific data server fails and a single copy fails, the distributed file system can recover the failed copy using the copy that has no failure.

Each of the data server 1 103-1, the data server 2 103-2, or the data server 3 103-3 constituting the distributed file system can manage the replica state data. In addition, the client server 101 and the metadata server 102 constituting the distributed file system may manage the replica state data.

At this time, since each of Data Server 1 (103-1), Data Server 2 (103-2), or Data Server 3 (103-3) stores a separate copy, the distributed file system is a replica managed by each component. By matching state data to be identical to each other, it is possible to maintain consistency of the files the replicas group and manage. The process of maintaining file consistency may be performed through copy state matching between the client server 101 and the metadata server 102 during the file operation process.

3 is a diagram illustrating a process of switching a replica state according to an embodiment of the present invention.

The distributed file system according to an embodiment of the present invention may divide and handle a failure occurring in the data server 103 into a temporary failure and a permanent failure. At this time, the metadata server 102 may monitor the status of the data servers 103 constituting the distributed file system, and determine the type of failure as a temporary failure or a permanent failure.

In this case, the temporary failure refers to a failure that returns to normal within a short time after being caused by an external factor (for example, a network failure) rather than a defect of the data server itself. In contrast, a permanent failure is a failure that occurs because of a fault in the data server itself (eg, a power supply failure) that prevents normal operation.

Referring to FIG. 3, when the replica state is N (Normal), the data server 103 means a normal state without a failure. If the failure of the data server 103 is detected, the replica status is F (Failure).

After that, when the temporary failure is eliminated and the data server 103 performs replica recovery, the replica status becomes R (Recovery). At this time, if a temporary failure occurs again in the data server 103, the replica state returns to F. FIG. And, when the restoration of the replica is completed, the replica state can be switched from R to N.

If the replica status is F, and the occurrence of a permanent failure is detected, the replica move begins and the replica status switches to M (Migration). If the replica fails, the replica status can be switched from M to F. And, when the movement of the replica is completed, the replica status can be switched from M to N.

At this time, when the replica is in the F state, a Missed Operation (MO) log for a file operation may be generated. In this case, the MO log may refer to a log in which a file operation content requested by the client server 101 to the data server 103 is recorded during a failure period from a temporary failure occurrence to a temporary failure resolution time.

In addition, when the replica is in the R state or the M state, a delayed operation (DO) log for a file operation may be generated. In this case, the DO log may mean a log in which a file operation content requested by the client server 101 to the data server 103 is recorded when the replica is being restored or the replica is moving.

In the case of a temporary failure, the data server 103 may finally recover the replica by replaying the MO log, and replay the DO log by performing a delayed file operation after the replica recovery ends. In the case of a permanent failure, the data server 103 may finally handle the failure by moving the replica, replaying the DO log and performing a delayed file operation after the completion of the replica move.

For example, each of the client server 101, the metadata server 102, and the plurality of data servers 103 constituting the distributed file system may store and maintain replica state data that they recognize. In this case, it may be assumed that the data server 103 maintains replica status data indicating a replica status of a replica managed by itself and a replica status of a replica identical to the replica managed by other data servers. If a failure occurs in the data server 103, significant overhead may occur because all replicas managed by the plurality of data servers 103 must be searched for failure processing.

Accordingly, according to an embodiment of the present invention, each data server may store and maintain replica state data indicating a replica state of a replica managed by itself and a server state of other data servers storing the same replica as the replica. At this time, the server state may mean that the replica state of the replicas stored by the other data servers is replaced. In addition, the metadata server 102 can also store and maintain replica status data indicating the server status of all data servers 103 that are storing replicas.

According to an embodiment of the present invention, the distributed file system may update the replica state data when the state of the replica is switched to maintain file consistency throughout the system. In other words, since file operation requests for a replica are sent by referring to the replica status data, file consistency is not maintained between data servers storing the same replica when the replica status data managed by the components of the distributed file system is different. This may occur.

This problem may occur because the client server 101 is a user level server, and it is difficult to immediately update the state of the replica. Therefore, the distributed file system can maintain file coherency through replica state matching during file operations.

4 is a diagram illustrating a process of performing a file operation by using a copy state matching in a distributed file system according to an exemplary embodiment of the present invention.

Referring to FIG. 4, the data server 1 103-1, the data server 2 103-2, and the data server 3 103-3 may store the same owner-specific replica X. FIG. The client server 101 may store and maintain the replica status data 401, and the metadata server 102 may store and maintain the replica status data 402. In addition, data server 1 103-1 indicates replica state data 403, data server 2 103-2 indicates replica state data 404, and data server 3 103-3 indicates replica state data ( 405 may be stored and maintained.

In one example, the data server may store and manage replica status data indicating the replica status of the replica it stores and the server status of the data server storing the same replica as the replica. Referring to FIG. 4, in the case of the replica state data 401, the state of the replica managed by the data server 1 103-1 is normal, and the server state of the data server 2 103-2 is normal. The server state of the data server 3 103-3 may be recognized as a recovery state.

4 shows an example in which a file operation is rejected by copy state matching. Specifically, FIG. 4 illustrates a state in which a temporary failure occurs in the data server 3 103-3 and the copy stored in the data server 3 103-3 is in a F state, and then the temporary failure is resolved to restore the copy. At this time, Figure 4 shows that the data server 3 (103-3) is returned to the normal state after the temporary failure is resolved, the replica to the metadata server 102 and the data servers (103-1, 103-2, 103-3) The situation after notifying that the state has been changed from the F state to the R state is shown. At this time, since the client server 101 has not sent a file operation request after the replica has changed to the R state, the replica status data 401 recognized by the client server 101 is still (N, N, F). .

Specifically, it may be difficult for the client server 101 to immediately recognize that the server state of the data servers 103-1, 103-2, and 103-3 is changed in correspondence to a user level server. That is, the client server 101 may not update the state changes of the data servers 103-1, 103-2, and 103-3 to the replica state data 403 quickly.

Then, the client server 101 changes the replica to the data server 1 103-1 and the data server 2 103-2 except for the data server 3 103-3 which is in the F state according to the replica status data 401. A file operation request can be sent. As a result, data server 1 103-1 and data server 2 103-2 store the changed copy according to the file operation request, while data server 3 103-3 stores the unchanged copy. In this case, file consistency is not maintained throughout the distributed file system.

Therefore, the distributed file system according to an embodiment of the present invention can maintain the file consistency of the entire system through matching the replica state during the file operation. Hereinafter, a process of maintaining file consistency through replica state matching will be described.

Referring to FIG. 4, the client server 101 includes the data server 1 103-1 and the data server 2 103-2 except for the data server 3 103-3 in the F state according to the replica state data 401. You can send a file operation request to change the replica. At this time, the client server 101 may include the replica status data 401 in the file operation request.

Then, the data server 1 103-1 may determine whether the replica status data 401 and the replica status data 403 match. The data server 2 103-2 may also determine whether the replica status data 401 and the replica status data 404 match. However, since (N, N, F), which is the replica status data 401, and (N, N, R), which is the replica status data 403, do not match, the data server 1 103-1 is the client server 101. You can reject the file operation request of). The same applies to the data server 2 103-2.

The client server 101 can then update the replica status data 401 to be the same as the replica status data 402 stored in the metadata server 102. That is, the client server 101 may change (N, N, F), which is the replica status data 401, to be the same as (N, N, R), which is the replica status data 402.

Here, the metadata server 102 periodically stores the replica status managed by each of the data server 1 103-1, the data server 2 103-2, and the data server 3 103-3 to the replica status data 402. Can be reflected.

Specifically, the metadata server 102 copies the server state in which the replica state managed by each of the data server 1 103-1, the data server 2 103-2, and the data server 3 103-3 is replaced. Status data 402 can be configured. At this time, the metadata server 102 may detect the state of the data servers in real time, and reflect the state of the current data server in the replica state data 402.

The client server 101 then issues a file operation request to the data servers 103-1, 103-2, and 103-3 that includes the updated replica state data 401 (N, N, R). You can send it again. At this time, the data servers 103-1, 103-2, and 103-3 may again determine whether the replica state data matches.

At this time, if the replica status data match each other, each of the data servers 103-1, 103-2, and 103-3 may perform a file operation, and if it does not match, may reject the file operation request again. . The process can be repeated until the replica status data matches each other.

Therefore, the distributed file system according to an embodiment of the present invention can maintain the file consistency of the entire system through matching the replica state during the file operation process.

5 is a flowchart illustrating a file consistency management method using copy state matching according to an embodiment of the present invention from the perspective of a data server.

The data server 103 described in FIG. 5 refers to any one data server 103 of the plurality of data servers 103 constituting the distributed file system according to an embodiment of the present invention.

In step S501, the data server 103 may store and maintain the replica and the first replica status data. In this case, the replica may mean a set of files grouped by owner. The owner may correspond to a directory to which the file belongs.

In one example, the first replica status data may represent a replica status of a replica and a server status of each of the other data servers 103 storing the same replica as the replica. At this time, the server state may mean that the replica state for the replica stored in each of the other data servers 103 is replaced.

For example, referring to FIG. 4, data server 1 103-1 is a replica state N for replica X and a server state N of data server 2 103-2 and data server 3 103-3. Replica status data 403 may be stored. At this time, the server state N of the data server 2 (103-2) and the data server 3 (103-3) of the replica X stored in the data server 2 (103-2) and the data server 3 (103-3) Replica status has been replaced.

In step S502, the data server 103 may receive a file operation request including the second copy state data from the client server 101. The second replica status data refers to replica status data stored by the client server 101.

In operation S503, the data server 103 may determine whether the first replica state data and the second replica state data match. At this time, when the first copy state data and the second copy state data match, in step S504, the data server 103 may perform a file operation in response to the file operation request.

On the other hand, when the first replica state data and the second replica state data do not match, in step S505, the data server 103 may reject the file operation request. In step S506, the data server 103 may re-receive the file operation request including the updated second replica status data from the client server 101.

At this time, the client server 101 may update the second replica state data to be the same as the third replica state data stored in the metadata server 102. Here, the third replica state data may be configured as current server states of the data servers 103 connected to the metadata server 102. That is, the metadata server 102 may store third copy state data reflecting the latest state of the data servers 103. At this time, the current server state of the data servers 103 connected to the metadata server 102 may be replaced by the replica state stored in the data server 103.

Then, in step S503, the data server 103 may again determine whether the first replica state data matches the updated second replica state data. At this time, if the first replica status data and the updated second replica status data match, step S504 is progressed, otherwise steps S505 and 506 are repeatedly performed.

6 is a flowchart illustrating a file consistency management method using copy state matching according to an embodiment of the present invention from the perspective of the client server 101. For reference, the first replica state data and the second replica state data described in FIG. 5 are separate from the description of FIG. 6.

In step S601, the client server 101 may store and maintain the first replica state data.

In operation S602, the client server 101 may transmit a fine operation request including the first copy state data to the data server 103.

In operation S603, when the first replica status data and the second replica status data stored in the data server 103 match, the client server 101 may receive a file operation result from the data server 103. At this time, the data server 103 may store and maintain a replica for each owner and second replica status data indicating the status of the replica. In this case, the replica may mean a set of files grouped by owner. The owner may correspond to a directory to which the file belongs. In one example, the data server 103 may determine whether the first replica status and the second replica status data match. When the first replica state and the second replica state match, the data server 103 may perform a file operation and return the file operation result to the client server 101.

In one example, the second replica status data may represent a replica status of a replica per owner and a server status of each of the other data servers 103 storing the same replica as the replica per owner. At this time, the server state may mean that the replica state of the replica stored by each of the other data servers 103 storing the same replica as the owner-specific replica is replaced.

In operation S604, when the first replica state data and the second replica state data do not match, the client server 101 may receive a rejection response to the file operation request from the data server 103. In operation S605, the client server 101 may update the first replica state data.

In detail, in operation S605, the client server 101 may update the first replica state data to be the same as the third replica state data stored in the metadata server 102. In this case, the third replica state data may be configured as a current server state of the data server 103 connected to the metadata server 102. That is, the metadata server 102 may update the third copy state data to the latest server state of the data server 103.

Then, in step S602, the client server 101 may retransmit the file operation request including the updated first copy state data to the data server 103. Then, the process may proceed to step S603 or S604 according to a matching result between the first copy state data and the second copy state data. That is, steps S604 and S605 may be repeatedly performed until the first copy state data and the second copy state data match.

7 is a block diagram showing the overall configuration of a data server performing a file consistency management method using replica state matching according to an embodiment of the present invention.

Referring to FIG. 7, the data server 103 may include a data storage unit 701, a file operation request receiver 702, a data matching determination unit 703, and a file operation execution unit 704.

The data storage unit 701 may store and maintain the replica and the first replica status data. In this case, the replica may mean a set of files grouped by owner. The owner may correspond to a directory to which the file belongs.

In one example, the first replica status data may represent the replica status of the replica and the server status of each of the other data servers storing the same replica as the replica. At this time, the server state may mean that the replica state for the replica stored in each of the other data servers is replaced.

The file operation request receiver 702 may receive a file operation request including the second copy state data from the client server 101.

The data matching determination unit 703 may determine whether the first replica state data and the second replica state data match.

The file operation performing unit 704 may perform a file operation when the first replica state data and the second replica state data match. For example, the file operation performing unit 704 may reject the file operation request of the client server 101 when the first replica state data and the second replica state data do not match.

Then, the client server 101 may update the second replica state data to be the same as the third replica state data stored in the metadata server 102. In this case, the third replica state data may be configured as a current server state of the data server connected to the metadata server 102. And,

In one example, the file operation request receiver 702 may re-receive a file operation request including the updated second copy state data from the client server 101. Accordingly, the data matching determination unit 703 may determine whether the first replica state data and the updated second replica state data match.

8 is a block diagram showing the overall configuration of a client server 101 performing a file consistency management method using replica state matching according to an embodiment of the present invention.

Referring to FIG. 8, the client server 101 may include a data storage 801, a file operation request transmitter 802, a data updater 803, and a file operation result receiver 804.

The data storage unit 801 may store and maintain the first copy state data.

The file operation request transmitter 802 may transmit a file operation request including the first copy state data to the data server 103. In this case, when the file operation contents change the replica, the file operation request transmitter 802 may transmit a file operation request to all data servers 103 storing the replica.

As an example, the data server 103 may store and maintain a replica, which is a collection of files grouped by owner, and second replica status data indicating a status of the replica. At this time, the second replica status data may represent the replica status of the replica and the server status of each of the other data servers storing the same replica as the replica. Server state means that the replica state stored by other data servers storing the same replica is replaced.

The data updater 803 may update the first replica state data to be the same as the third replica state data stored in the metadata server. For example, if the first replica state data and the second replica state data do not match, the data updater 803 may update the first replica state data to be the same as the third replica state data. In this case, the third replica state data may be configured as a current server state of the data server 103 connected to the metadata server 103.

The file operation result receiver 804 may receive a file operation result from the data server 803 when the first replica state data and the second replica state data stored in the data server 103 match. In this case, when the first replica state data and the second replica state data do not match, the file operation result receiver 804 may receive a rejection response to the file operation request from the data server 103.

As mentioned above, the data updater 803 may update the first replica status data to be the same as the third replica status data of the metadata server 102 according to the rejection response. Then, the file operation request transmitter 801 may retransmit the file operation request including the updated first copy state data to the data server 103.

In addition, the method of managing file consistency using copy state matching according to an embodiment of the present invention includes a computer readable medium including program instructions for performing various computer-implemented operations. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The media may be program instructions that are specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

As described above, the present invention has been described by way of limited embodiments and drawings, but the present invention is not limited to the above-described embodiments, which can be variously modified and modified by those skilled in the art to which the present invention pertains. Modifications are possible. Accordingly, the spirit of the present invention should be understood only by the claims set forth below, and all equivalent or equivalent modifications thereof will belong to the scope of the present invention.

1 is a view showing the overall configuration of a distributed file system according to an embodiment of the present invention.

2 is a diagram illustrating an example in which files are grouped in units of owners based on directories according to an embodiment of the present invention, and copies of owners are stored in a data server.

3 is a diagram illustrating a process of switching a replica state according to an embodiment of the present invention.

4 is a diagram illustrating a process of performing a file operation by using a copy state matching in a distributed file system according to an exemplary embodiment of the present invention.

5 is a flowchart illustrating a file consistency management method using copy state matching according to an embodiment of the present invention from the perspective of a data server.

6 is a flowchart illustrating a file consistency management method using replica state matching according to an embodiment of the present invention from the perspective of a client server.

7 is a block diagram showing the overall configuration of a data server performing a file consistency management method using replica state matching according to an embodiment of the present invention.

8 is a block diagram showing the overall configuration of a client server performing a file consistency management method using replica state matching according to an embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

101: client server

102: metadata server

103: data server

Claims (32)

The data server storing and maintaining replica and first replica state data; Receiving, by the data server, a file operation request including second replica state data from a client server; Determining, by the data server, whether the first replica state data and the second replica state data match; And The data server performing a file operation when the first replica state data and the second replica state data match. File consistency management method using replica status matching including a. The method of claim 1, The replica is, A collection of files grouped by owner, The owner, File consistency management method using replica status matching, characterized in that corresponding to the directory to which the file belongs. The method of claim 1, The first replica state data, And a server state of each of the other data servers storing the replica state of the replica and the same replica as the replica. The method of claim 3, The server state is, File consistency management method using replica status matching, characterized in that the replica status for the replica stored in each of the other data server is replaced. The method of claim 1, Rejecting, by the data server, a file operation request when the first replica state data and the second replica state data do not match; And The data server re-receiving a file operation request containing the updated second replica state data from the client server File consistency management method using replica status matching further comprising. The method of claim 5, The determining of whether the data server matches the first copy state data and the second copy state data may include: And determining whether the data server matches the first replica state data and the updated second replica state data. The method of claim 5, The client server, And updating the second replica state data to be the same as the third replica state data stored in the metadata server. The method of claim 7, wherein The third replica state data, And a current server state of a data server connected to the metadata server. Storing and maintaining, by the client server, the first replica state data; Sending, by the client server, a file operation request including the first replica state data to a data server; And Receiving a file operation result from the data server when the first replica state data and the second replica state data stored in the data server match. File consistency management method using replica status matching including a. 10. The method of claim 9, The data server, A method for managing file consistency using replica state matching, which comprises storing and maintaining a replica for each owner and second replica state data indicating a state of the replica. The method of claim 10, The replica is, A collection of files grouped by owner, The owner, File consistency management method using replica status matching, characterized in that corresponding to the directory to which the file belongs. The method of claim 10, The second replica state data, And a server state of each of the other data servers storing the replica state of the replica by the owner and the same replica as the replica by the owner. The method of claim 12, The server state is, And a replica state of a replica stored in each of the other data servers is replaced. 10. The method of claim 9, Receiving, by the client server, a rejection response to a file operation request from the data server when the first replica state data and the second replica state data do not match; And The client server updating the first replica status data File consistency management method using replica status matching further comprising. The method of claim 14, The client server transmitting a file operation request including the first replica state data to a data server, And the client server retransmits a file operation request including the updated first replica state data. The method of claim 14, Updating the first replica state data, And updating the first replica state data to be the same as the third replica state data stored in the metadata server. The method of claim 16, The third replica state data, And a current server state of a data server connected to the metadata server. A computer-readable recording medium in which a program for executing the method of any one of claims 1 to 17 is recorded. A data storage unit for storing and maintaining the replica and the first replica state data; A file operation request receiver configured to receive a file operation request including second copy state data from a client server; A data matching determination unit determining whether the first replica state data and the second replica state data match; And A file operation performing unit for performing a file operation by the data server when the first replica state data and the second replica state data match. Data server comprising a. The method of claim 19, The replica is, A collection of files grouped by owner, The owner, And a data server corresponding to a directory to which the file belongs. The method of claim 19, The first replica state data, A replica status of the replica and a server status of each of the other data servers storing the same replica as the replica. The method of claim 21, The server state is, And a replica state for the replica stored by each of the other data servers is replaced. The method of claim 19, The file operation execution unit, And if the first replica state data and the second replica state data do not match, rejecting the file operation request. The method of claim 19, The file operation request receiver, Re-receive a file operation request containing the updated second replica state data from the client server, The data matching determination unit, And determining whether the first replica state data matches the updated second replica state data. The method of claim 24, The client server, And update the second replica state data to be the same as the third replica state data stored in the metadata server. The method of claim 25, The third replica state data, And a current server state of a data server connected to the metadata server. A data storage unit for storing and maintaining first replica state data; A file operation request transmitter for transmitting a file operation request including the first copy state data to a data server; And A file operation result receiver configured to receive a file operation result from the data server when the first replica state data and the second replica state data stored in the data server match. Client server comprising a. The method of claim 27, The data server, And storing and maintaining a replica, which is a collection of files grouped by owner, and second replica status data representing the replica's status. The method of claim 28, The second replica state data, And a server state of each of the other data servers storing the replica state of the replica and the same replica as the replica. The method of claim 27, The file operation result receiver, And if the first replica state data and the second replica state data do not match, receiving a rejection response to a file operation request from the data server. The method of claim 27, A data updater for updating the first replica status data to be the same as the third replica status data stored in the metadata server More, The file operation request transmission unit, And resend a file operation request containing the updated first replica state data to the data server. The method of claim 31, wherein The third replica state data, And a current server state of a data server connected to the metadata server.

Images