JP2008097155A - Data storage controller and data storage device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve efficiency of data processing by suppressing increase of overhead related to transaction. <P>SOLUTION: A primary master transmits a request message to execute an operation. A primary cohort executes an operation and transmits an ACK message to the primary master when it receives the execution request message, and transmits a message including log information related to the operation, and a backup cohort receives the message and stores the log information in a memory. The primary master transmits a request message to execute decision processing after completion of all operations. The primary cohort, in receiving the execution request message, executes the decision processing and transmits a message including the log information related to the decision processing. The backup cohort receives the message and executes the decision processing based on the log information stored in the memory. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention has a data storage control device and a plurality of data storage units that perform control for storing data and backup data of the data in different data storage units, and the data and backup data of the data are different from each other. More particularly, the present invention relates to a data storage control device and a data storage device capable of suppressing an increase in overhead associated with a transaction and improving data processing efficiency.

  Conventionally, each disk connected via a network such as a SAN (Storage Area Network) performs autonomous data management using a processor and memory on a disk controller, and communicates between the disks from the host. An autonomous disk that executes requested data processing has been proposed (see Non-Patent Document 1).

  In this autonomous disk, in order to improve the reliability of data management, a backup of data stored in the disk is stored in a disk different from the disk. Then, when a failure occurs in the disk and data cannot be read, the backed up data is read instead and data processing is executed.

  In such a system, it is extremely important to ensure the atomicity of transactions. As a commit protocol for realizing this, there are a two-phase commit (2PC) protocol, an early prepare (ER) protocol, and the like (see Non-Patent Document 2).

Haruo Yokota, “Autonomous Disks for Advanced Database Applications”, in Proc. Of International Symposium on Database Applications in Non-traditional Environments (DANTE'99), pp.448-457. J. Stamos and F. Cristian, “A low-cost atomic commit protocol”, In proceeding of ninth symposium on Reliable distributed systems, October 1990.

  However, when the above-described commit protocol in the prior art is applied to an autonomous disk that performs processing for backing up data stored in a disk to another disk, there is a problem that satisfactory performance cannot be obtained.

  Specifically, in the two-phase commit protocol or the early prepare protocol, it is necessary to write a log to the disk device for each operation and return a message indicating that the writing has been completed normally to the device that requested the execution of the operation. As a result, the overhead associated with the transaction increases and the processing efficiency of the autonomous disk decreases.

  Thus, in an autonomous disk, how to suppress an increase in overhead related to a transaction and improve data processing efficiency is an important issue.

  The present invention has been made to solve the above-described problems caused by the prior art, and is a data storage control device and a data storage device capable of suppressing an increase in overhead associated with a transaction and improving data processing efficiency. The purpose is to provide.

  In order to solve the above-described problems and achieve the object, the present invention is a data storage control for performing control for storing data and backup data of the data in different data storage units (disk devices 40a to 40d in the embodiment). A first storage control unit (primary master in the embodiment) that transmits a plurality of operation execution request messages for data, and when the operation execution request message is received, executes the operation; Each time execution of the operation is completed, a response message to the operation execution request message is transmitted to the first storage control unit, log information related to the operation is generated, and a message including the log information is transmitted. Second storage control unit (in the example Primary cohort) and one or more third storage control units that receive messages transmitted by the one or more second storage control units and store log information related to each operation in a memory (Example) The first storage controller receives response messages sent by the one or more second storage controllers and whether or not all operations have been executed. After the completion of execution of all the operations, a decision processing execution message for the operation is transmitted, and the one or more second storage control units transmit the decision processing transmitted by the first storage control unit. When an execution request message is received, a decision process is executed and connected to each of the one or more second storage control units. Controlling update of data stored in the data storage unit, generating log information related to decision processing and transmitting a message including the log information, the one or more third storage control units, Each of the one or more third storage control units receives a message transmitted by one or more second storage control units and executes a decision process based on the log information stored in the memory. The update control of the backup data stored in the data storage unit connected to is performed.

  Further, the present invention is the above invention, wherein the one or more third storage control units, after executing the decision process, transmit a decision completion message indicating completion of the decision process, and When the two storage control units receive the decision completion message from the one or more third storage control units, the synchronization completion indicating that the data synchronization processing has been completed to the first storage control unit A message is transmitted.

  Further, according to the present invention, in the above invention, a fourth storage control unit (backup in the embodiment) for storing membership log information indicating one or more second storage control units for transmitting an operation execution request message in a memory. The first storage control unit generates the membership log information and transmits the membership log information to the fourth storage control unit. When the execution of the decision process is determined, the first storage control unit determines the execution of the decision process. Is generated and transmitted to the fourth storage control unit, and after receiving the decision log information, the fourth storage control unit detects a failure related to the first storage control unit A decision process execution request message for an operation based on the membership log information. And transmits the to the storage control unit.

  Further, the present invention is the above invention, wherein the one or more second storage control units receive the decision processing execution request message from the first storage control unit, and then the one or more third storage control units. When a failure related to the storage control unit is detected, the decision completion message is transmitted to the first storage control unit, and when the failure is recovered, the one or more third storage control units are: Executing a decision process based on the log information stored in the memory, and performing update control of backup data stored in a data storage unit connected to each of the one or more third storage control units; It is characterized by.

  Further, the present invention is the above invention, wherein the first storage control unit receives all response messages transmitted by the one or more second storage control units, and then receives the one or more second storage control units. When a failure related to any one of the storage control units is detected, a decision process execution request message for an operation is transmitted to the one or more third storage control units, and the one or more third storage control units are transmitted. The storage control unit executes the decision process when receiving the decision process execution request message transmitted by the first storage control unit, and recovers the log information related to the decision process from the failure when the failure is recovered The second storage control unit, which has transmitted to the second storage control unit and recovered from the failure, receives the log information related to the decision process, and determines based on the received log information. And executes the ® down process.

  Further, the present invention is a data storage device that has a plurality of data storage units (disk devices 40a to 40d in the embodiment) and stores data and backup data of the data in different data storage units, When a first storage control unit (primary master in the embodiment) for transmitting a plurality of operation execution request messages for the operation and the operation execution request message is received, the operation is executed and the execution of the operation is completed. One or more second storage control units each transmitting a response message to the operation execution request message to the first storage control unit, generating log information related to the operation, and transmitting a message including the log information (Primary cohort in the examples); One or more third storage control units (backup cohort in the embodiment) that receive messages transmitted by one or more second storage control units and store log information related to each operation in a memory; The first storage control unit receives the response message transmitted by the one or more second storage control units, determines whether or not execution of all operations is completed, After the execution of the operation is completed, a decision process execution request message for the operation is transmitted, and the one or more second storage control units receive the decision process execution request message transmitted by the first storage control unit In this case, a decision process is executed to store the data stored in the data storage unit connected to each of the one or more second storage control units. The log data relating to the decision process and transmitting a message including the log information, wherein the one or more third storage control units are configured to transmit the one or more second storage controls. Receiving a message transmitted by the unit, executing a decision process based on the log information stored in the memory, and storing the data in a data storage unit connected to each of the one or more third storage control units The backup data that has been updated is updated.

  Further, the present invention is the above invention, wherein the one or more third storage control units, after executing the decision process, transmit a decision completion message indicating completion of the decision process, and When the two storage control units receive the decision completion message from the one or more third storage control units, the synchronization completion indicating that the data synchronization processing has been completed to the first storage control unit A message is transmitted.

  Further, according to the present invention, in the above invention, a fourth storage control unit (backup in the embodiment) for storing membership log information indicating one or more second storage control units for transmitting an operation execution request message in a memory. The first storage control unit generates the membership log information and transmits the membership log information to the fourth storage control unit. When the execution of the decision process is determined, the first storage control unit determines the execution of the decision process. Is generated and transmitted to the fourth storage control unit, and after receiving the decision log information, the fourth storage control unit detects a failure related to the first storage control unit A decision process execution request message for an operation based on the membership log information. And transmits the to the storage control unit.

  Further, the present invention is the above invention, wherein the one or more second storage control units receive the decision processing execution request message from the first storage control unit, and then the one or more third storage control units. When a failure related to the storage control unit is detected, the decision completion message is transmitted to the first storage control unit, and when the failure is recovered, the one or more third storage control units are: A decision process is executed based on the log information stored in the memory to update backup data stored in a data storage unit connected to each of the one or more third storage control units. And

  Further, the present invention is the above invention, wherein the first storage control unit receives all response messages transmitted by the one or more second storage control units, and then receives the one or more second storage control units. When a failure related to any one of the storage control units is detected, a decision process execution request message for an operation is transmitted to the one or more third storage control units, and the one or more third storage control units are transmitted. The storage control unit executes the decision process when receiving the decision process execution request message transmitted by the first storage control unit, and recovers the log information related to the decision process from the failure when the failure is recovered The second storage control unit, which has transmitted to the second storage control unit and recovered from the failure, receives the log information related to the decision process, and determines based on the received log information. And executes the ® down process.

  The present invention is also a data storage method for storing data and backup data of the data in different data storage units (disk devices 40a to 40d in the embodiment), and includes a first storage control unit (in the embodiment). The operation execution request message transmission step in which the primary master) transmits an operation execution request message for data, and one or more second storage control units (primary cohort in the embodiment) receive the operation execution request message. In this case, the operation is executed, and a response message for the operation execution request message is transmitted to the first storage control unit every time the execution of the operation is completed, and log information related to the operation is generated to generate the log information. Messe containing Log information concerning each operation by receiving one or more third storage control units (the backup cohort in the embodiment) receiving the message transmitted by the one or more second storage control units. And an operation execution step of storing all the operations in the memory, and whether or not the first storage control unit has received the response message transmitted by the one or more second storage control units and has completed the execution of all the operations. A decision execution request message transmission step for transmitting a decision processing execution message for the operation after completion of execution of all the operations, and the one or more second storage control units include a first storage control The decision processing is executed when the decision processing execution request message transmitted by the unit is received, Updating data stored in the data storage unit connected to each of the two or more second storage control units, generating log information related to the decision processing and transmitting a message including the log information, One or more third storage control units receive the message transmitted by the one or more second storage control units, execute a decision process based on the log information stored in the memory, A decision execution step of updating backup data stored in a data storage unit connected to each of the one or more third storage control units.

  Further, the present invention is the above-mentioned invention, wherein the one or more third storage control units execute a decision process, and then transmit a decision completion message indicating completion of the decision process. When the one or more second storage control units receive a decision completion message from the one or more third storage control units, the data synchronization processing is completed for the first storage control unit And a synchronization completion message transmission step of transmitting a synchronization completion message indicating that the operation has been completed.

  Further, the present invention is the above invention, wherein the operation execution request message transmission step includes a membership in which the first storage control unit indicates one or more second storage control units that transmit an operation execution request message. Log information is generated and transmitted to a fourth storage control unit (backup master in the embodiment), the fourth storage control unit stores the membership log information in a memory, and the decision execution request message transmission step includes: When the first storage control unit decides to execute the decision process, it generates decision log information indicating the decision to execute the decision process and transmits it to the fourth storage control unit. If the failure relating to the first storage control unit is detected after receiving the decision log information, the membership And transmitting the decision process of the execution request message for operation said one or more second storage controller based on the grayed information.

  Further, the present invention is the above-described invention, wherein the decision execution step is performed after the one or more second storage control units receive the decision processing execution request message from the first storage control unit. When a failure related to one or more third storage control units is detected, the decision completion message is transmitted to the first storage control unit, and when the failure is recovered, the one or more third storage control units are transmitted. Backup data stored in a data storage unit connected to each of the one or more third storage control units by executing a decision process based on the log information stored in the memory. It is characterized by updating.

  Also, in the present invention according to the above invention, in the decision execution request message transmission step, the first storage control unit has received all response messages transmitted by the one or more second storage control units. Then, when a failure related to any one of the one or more second storage control units is detected, a decision processing execution request message for an operation is transmitted to the one or more third storage control units. The decision execution step executes the decision process when the one or more third storage control units receive a decision process execution request message transmitted by the first storage control unit, and When log is recovered, log information related to decision processing is transmitted to the second storage control unit recovered from the failure, and the second storage control unit recovered from the failure is It receives the log information related to decision processing, and executes the decision processing based on the received log information.

  Further, the present invention is a data storage control program for performing control for storing data and backup data of the data in different data storage units (disk devices 40a to 40d in the embodiment), the first storage control unit The operation execution request message transmission procedure in which the (primary master in the embodiment) transmits an operation execution request message for data, and one or more second storage control units (primary cohort in the embodiment) execute the operation. When the request message is received, the operation is executed, and a response message to the operation execution request message is transmitted to the first storage control unit every time the execution of the operation is completed, and log information related to the operation is generated. Tote A message including log information is transmitted, and one or more third storage control units (backup cohorts in the embodiment) receive the messages transmitted by the one or more second storage control units, and An operation execution procedure for storing log information related to an operation in a memory, and execution of all operations by the first storage control unit receiving a response message transmitted by the one or more second storage control units A decision execution request message transmission procedure for transmitting a decision processing execution message for an operation after completion of execution of all operations, and the one or more second storage control units, When a decision processing execution message sent by the first storage control unit is received, the decision is made Update the data stored in the data storage unit connected to each of the one or more second storage control units, and generate log information related to the decision process to generate the log information. And the one or more third storage control units receive the message transmitted by the one or more second storage control units, and log information stored in the memory is stored in the log information. Executing a decision process based on the decision execution procedure for updating backup data stored in a data storage unit connected to each of the one or more third storage control units. It is characterized by.

  According to the present invention, when the first storage control unit transmits an operation execution request message for data and one or more second storage control units receive the operation execution request message, the operation is performed. Each time execution of the operation is completed, a response message to the operation execution request message is transmitted to the first storage control unit, log information relating to the operation is generated, and a message including the log information is transmitted. The one or more third storage control units receive the message transmitted by the one or more second storage control units, store the log information related to each operation in the memory, and perform the first storage control. The unit receives response messages sent by one or more second storage controllers and executes all operations It is determined whether or not it has been completed, and after execution of all operations is completed, a decision processing execution request message for the operation is transmitted, and one or more second storage control units are transmitted by the first storage control unit When the decision processing execution request message is received, the decision processing is executed to update the data stored in the data storage unit connected to each of the one or more second storage control units, and the decision processing Generating log information and transmitting a message including the log information, wherein one or more third storage control units receive a message transmitted by one or more second storage control units, A data storage unit that executes decision processing based on log information stored in the memory and is connected to each of the one or more third storage control units Since it was decided to update the stored backup data, suppressing an increase in the overhead of the transaction, it is possible to improve the efficiency of data processing.

  Further, according to the present invention, after one or more third storage control units execute a decision process, a decision completion message indicating completion of the decision process is transmitted, and one or more second storage control units However, when a decision completion message is received from one or more third storage control units, a synchronization completion message indicating that data synchronization processing has been completed is sent to the first storage control unit Therefore, it is possible to ensure synchronization between the data and the backup data.

  Further, according to the present invention, the first storage control unit generates membership log information indicating one or more second storage control units that transmit the operation execution request message to generate the fourth storage control unit. The fourth storage control unit stores the membership log information in the memory, and when the first storage control unit determines the execution of the decision process, the decision log information indicating the execution determination of the decision process is displayed. Based on the membership log information when the fourth storage control unit detects a failure related to the first storage control unit after receiving the decision log information. The decision processing execution request message for the operation is transmitted to one or more second storage control units, so that a failure related to the first storage control unit occurs. , It is possible to quickly execute the decision process.

  Further, according to the present invention, one or more second storage control units receive one or more third storage control units after receiving a decision processing execution request message from the first storage control unit. When a failure is detected, a decision completion message is transmitted to the first storage control unit, and when the failure is recovered, one or more third storage control units store log information stored in the memory. Since the decision processing is executed based on this and the backup data stored in the data storage unit connected to each of the one or more third storage control units is updated, the third storage control unit Even when a failure occurs, the decision process can be executed quickly.

  In addition, according to the present invention, after the first storage control unit receives all response messages transmitted by the one or more second storage control units, the one or more second storage control units When a failure related to any one of them is detected, a decision processing execution message for the operation is transmitted to one or more third storage control units, and one or more third storage control units When the decision processing execution request message transmitted by the storage control unit is received, the decision processing is executed, and when the failure is recovered, the log information related to the decision processing is restored to the second storage control unit that has recovered from the failure. Since the second storage control unit that has transmitted and recovered from the failure has received the log information related to the decision processing, and decided to execute the decision processing based on the received log information. Even if a failure of the second storage controller has occurred, it can quickly perform the decision process.

  According to the present invention, since it is possible to improve the efficiency of data processing by suppressing an increase in overhead associated with a transaction, it is possible to speed up the response to a data processing request from a host, and an autonomous distributed storage such as an autonomous disk There is an effect that it is possible to effectively cope with an increase in storage nodes in the system.

  Exemplary embodiments of a data storage control device and a data storage device according to the present invention will be described below in detail with reference to the accompanying drawings.

  First, the functional configuration of the autonomous disk system according to the present invention will be described. FIG. 1 is a diagram illustrating a functional configuration of the autonomous disk system according to the present embodiment. As shown in FIG. 1, this autonomous disk system has a configuration in which clients 10a to 10c and processing elements 20a to 20d are connected via a network 50 such as a SAN (Storage Area Network).

  The clients 10a to 10c are client devices that transmit execution requests for various data processing such as data search, insertion, and deletion to the processing elements 20a to 20d.

  The processing elements 20a to 20d are processing devices that receive various data processing execution requests transmitted by the clients 10a to 10c and execute the requested data processing. Each processing element 20a to 20d stores data to be processed in a distributed manner. Each processing element 20a to 20d searches for data using a directory structure called Fat-Btree. The present invention is not particularly limited to the Fat-Btree method as a directory method, and can be applied to all distributed directory methods in which a directory is shared by a plurality of processing elements.

  FIG. 2 is an explanatory diagram for explaining Fat-Btree. Fat-Btree is an improvement over the conventional parallel Btree, and has a directory structure developed so that data can be accessed at high speed in a distributed database system such as an autonomous disk.

  In this Fat-Btree, as shown in FIG. 2, B-tree data pages are stored evenly in the processing elements 20a to 20d, and only the index pages that are higher than the data pages are processed elements 20a to 20d. Remember me. That is, in Fat-Btree, each processing element 20a to 20d stores a B-tree subtree.

  As described above, in Fat-Btree, an index page that is closer to the root page and has a higher access frequency is copied to more processing elements 20a to 20d, so that the search speed can be increased.

  Further, the lower the index page having the higher update frequency, the smaller the number of processing elements 20a to 20d that store the copy, and therefore the smaller the number of processing elements 20a to 20d that need to be synchronized when updating the copy. And overhead can be reduced.

  Returning to the description of FIG. 1, the processing elements 20a to 20d include controllers 30a to 30d and disk devices 40a to 40d.

  The controllers 30a to 30d are processing units that control data access to the disk devices 40a to 40d. The controllers 30a to 30d search for data using the Fat-Btree shown in FIG. 2, and execute various data processing on the data stored in the disk devices 40a to 40d. A detailed functional configuration of the controllers 30a to 30d will be described later with reference to FIG.

  The disk devices 40a to 40d are storage devices such as hard disk devices. Specifically, the disk devices 40a to 40d include the Fat-Btree directory information shown in FIG. 2, backup data of directory information stored in the other disk devices 40a to 40d, and data to be subjected to various data processing. The backup data of the data stored in the other disk devices 40a to 40d is stored.

  Next, the functional configuration of the processing elements 20a to 20d shown in FIG. 1 will be described in detail. FIG. 3 is a diagram showing a functional configuration of the processing elements 20a to 20d shown in FIG. The functional configuration of each processing element 20a to 20d is the same.

  As described with reference to FIG. 1, the processing element 20a (20b, 20c, 20d) includes a controller 30a (30b, 30c, 30d) and a disk device 40a (40b, 40c, 40d). The controller 30a (30b, 30c, 30d) includes a network interface unit 31a (31b, 31c, 31d), a disk interface unit 32a (32b, 32c, 32d), a memory 33a (33b, 33c, 33d), and a control unit 34a. (34b, 34c, 34d).

  The network interface unit 31a (31b, 31c, 31d) is a network interface that exchanges data with the clients 10a to 10c via the network 50. The disk interface unit 32a (32b, 32c, 32d) is a disk interface that exchanges data with the disk device 40a (40b, 40c, 40d).

  The memory 33a (33b, 33c, 33d) is a semiconductor main storage device in which data is read and written by the control unit 34a (34b, 34c, 34d). The memory 33a (33b, 33c, 33d) includes various data such as a log 330a (330b, 330c, 330d), backup disk management data 331a (331b, 331c, 331d), and ACK management data 332a (332b, 332c, 332d). Remember.

  The log 330a (330b, 330c, 330d) is log data storing information related to the contents of a series of operations performed in each transaction. The backup disk management data 331a (331b, 331c, 331d) is data that stores the correspondence between the disk devices 40a to 40d that store the original data and the disk devices 40a to 40d that store the backup data of the original data. is there.

  FIG. 4 is a diagram showing an example of the backup disk management data 331a (331b, 331c, 331d) shown in FIG. As shown in FIG. 4, this backup disk management data 331a (331b, 331c, 331d) stores a primary disk ID and a backup disk ID.

  The primary disk ID is identification information for identifying the disk devices 40a to 40d that store the original data. The backup disk ID is identification information for identifying the disk devices 40a to 40d that store the backup data of the original data.

  Returning to the description of FIG. 3, the ACK management data 332a (332b, 332c, 332d) is obtained when the control unit 34a (34b, 34c, 34d) transmits an operation execution request to the other processing elements 20b to 20d. Further, it is data indicating whether or not an ACK message indicating the completion of the operation has been received from the other processing elements 20b to 20d.

  FIG. 5 is a diagram illustrating an example of the ACK management data 332a (332b, 332c, 332d) illustrated in FIG. As shown in FIG. 5, the ACK management data 332a (332b, 332c, 332d) stores an operation ID, an operation target disk ID, and an ACK flag.

  The operation ID is identification information for identifying each operation that has transmitted an execution request to the other processing elements 20b to 20d. The operation target disk ID is identification information for identifying the disk devices 40b to 40d included in the processing elements 20b to 20d that have transmitted the operation execution request.

  The ACK flag is a flag indicating whether or not an ACK message has been received from the processing elements 20b to 20d that have transmitted the operation execution request. When this flag is “1”, it indicates that an ACK message has been received from the processing elements 20b to 20d. When this flag is “0”, it indicates that no ACK message has been received from the processing elements 20b to 20d. Show.

  Returning to the description of FIG. 3, the control unit 34 a (34 b, 34 c, 34 d) is a control unit that totally controls the controller 30 a (30 b, 30 c, 30 d), and controls data exchange between the functional units. The control unit 34a (34b, 34c, 34d) includes a transaction execution unit 340a (340b, 340c, 340d) and a failure recovery processing unit 341a (341b, 341c, 341d).

  When the transaction execution unit 340a (340b, 340c, 340d) receives an execution request for various data processing related to a transaction from the clients 10a to 10c or the other processing elements 20b to 20d, the transaction execution unit 340a (340b, 340c, 340d) is based on the Fat-Btree directory information. It is a functional unit that searches for data to be processed and executes various data processing on the data.

  The transaction execution unit 340a (340b, 340c, 340d) executes a transaction using the backup assist 1.5 phase commit protocol (BA-1.5PC). This BA-1.5PC will be described in detail later.

  The failure recovery processing unit 341a (341b, 341c, 341d) is a functional unit that performs recovery processing for a failure when a failure occurs in any of the processing elements 20a to 20d during execution of the transaction.

  The failure recovery processing unit 341a (341b, 341c, 341d) stores the role of the processing element 20a-20d in which the failure has occurred (whether the processing element 20a-20d in which the failure has occurred stores the original data or backup data). The recovery process is executed differently depending on whether it is something. This recovery process will also be described in detail later.

  The disk device 40a (40b, 40c, 40d) includes directory data 400a (400b, 400c, 400d), directory backup data 401a (401b, 401c, 401d), primary data 402a (402b, 402c, 402d), and backup data 403a ( 403b, 403c, 403d) are stored.

  Directory data 400a (400b, 400c, 400d) is data of a directory configured by Fat-Btree. The directory data 400a (400b, 400c, 400d) is data corresponding to the B-tree subtree shown in FIG.

  The directory backup data 401a (401b, 401c, 401d) is backup data of the directory data stored by the disk devices 40b to 40d of the other processing elements 20b to 20d.

  The primary data 402a (402b, 402c, 402d) is data that is an execution target of various data processing in a transaction. The backup data 403a (403b, 403c, 403d) is backup data of primary data stored by the disk devices 40b-40d of the other processing elements 20b-20d.

  Next, the asynchronous neighbor-WAL protocol will be described. FIG. 6 is a sequence diagram illustrating the asynchronous neighbor-WAL protocol. In this embodiment, an asynchronous neighbor-WAL protocol is used to suppress transaction log overhead.

  In this asynchronous neighbor-WAL protocol, when the primary processing element that stores the original data receives a log flush request (step S101), the log is written in the memory provided in the controller of the primary processing element (step S102).

  Further, the primary processing element transmits a log message including information on the log written in the memory to the backup processing element that stores the backup data (step S103).

  When the backup processing element receives this log message, the backup processing element writes the log in the memory provided in the controller of the backup processing element (step S104), and ends the log writing process using the asynchronous neighbor-WAL protocol.

  In this asynchronous neighbor-WAL protocol, the backup processing element does not send an ACK message indicating the completion of the log writing process to the primary processing element, so log writing is not synchronized. In addition, since the log is not written in the disk device but in the memory, the log overhead can be greatly suppressed. On the other hand, a case where an ACK message is transmitted from the backup processing element to the primary processing element to synchronize log writing is referred to as a synchronous neighbor-WAL protocol.

  Next, the BA-1.5 phase commit protocol according to this embodiment will be described. FIG. 7 is a sequence diagram illustrating the BA-1.5 phase commit protocol according to the present embodiment.

  In FIG. 7, the processing elements 20a to 20d have different roles of a primary master, a backup master, a primary cohort, and a backup cohort, respectively.

  The primary master is a processing element that receives a transaction execution request from the clients 10a to 10c and transmits execution requests for various operations in the transaction to the other processing elements 20a to 20d. The backup master is a processing element that acquires various logs related to transactions from the primary master and stores them for backup.

  The primary cohort is a processing element that stores original data that is updated in a transaction. The backup cohort is a processing element that stores backup data of data stored by the primary cohort.

  In the following description, it is assumed that the primary master, the backup master, the primary cohort, and the backup cohort are the processing element 20a, the processing element 20b, the processing element 20c, and the processing element 20d, respectively. It is assumed that there are not only one processing element 20c (primary cohort) and processing elements 20d (backup cohort), but also a plurality of processing elements.

  First, when the transaction execution unit 340a of the processing element 20a receives a transaction execution request from the clients 10a to 10c, the transaction execution unit 340a generates a membership log, stores it as the log 330a in the memory 33a, and processes the membership log. The information is transmitted to the element 20b, and the membership log is stored as the log 330b in the memory 33b of the processing element 20b (step S201).

  Here, the membership log is information of the processing element 20c that stores target data of various operations executed in the transaction, and is generated based on the directory data 400a.

  The transaction execution unit 340a of the processing element 20a determines the processing element 20b to which the membership log is to be transmitted by referring to the backup disk management data 331b, and transmits the membership log to the processing element 20b.

  The transaction execution unit 340b of the processing element 20b receives the membership log from the processing element 20a, and stores the membership log as a log 330b in the memory 33b.

  The transaction execution unit 340a of the processing element 20a transmits the membership log to the processing element 20b, and then transmits an operation execution request message related to the transaction to the plurality of processing elements 20c (step S202). Here, the processing element 20c that transmits the operation execution request message is specified by the membership log.

  The transaction execution unit 340c of each processing element 20c receives this operation execution request message and executes the operation. Then, the transaction execution unit 340c generates a log related to the operation, stores the log as the log 330c in the memory 33c using the asynchronous neighbor-WAL protocol described in FIG. 6, and backs up each processing element 20c. The log is transmitted to the processing element 20d that stores data (step S203). The transaction execution unit 340d of the processing element 20d that has received this log stores the transmitted log as the log 330d in the memory 33d.

  When the execution of the operation is completed, the transaction execution unit 340c of the processing element 20c transmits an ACK message indicating that the execution of the operation is completed to the processing element 20a that has transmitted the operation execution request message (step S204). .

  The transaction execution unit 340a of the processing element 20a receives the ACK message from the transaction execution unit 340c of the processing element 20c, and updates the ACK flag of the ACK management data 332a to “1”.

  Then, when there is an operation for requesting the processing element 20c to execute, the transaction execution unit 340a transmits an execution request message for the next operation to the processing element 20c (step S205).

  Hereinafter, the transaction execution unit 340a of the processing element 20a, the transaction execution unit 340c of the processing element 20c, and the transaction execution unit 340d of the processing element 20d perform the same processing as in steps S202 to S204 until all operations are completed. Repeat (step S206 to step S210).

  Then, the transaction execution unit 340a of the processing element 20a confirms that the ACK message corresponding to all the operations has been received from all the processing elements 20c registered in the membership log, and then commit processing (or abort processing) for the transaction. ) Is stored as a log 330a in the memory 33a, and the decision log is transmitted to the processing element 20b (step S211).

  The transaction execution unit 340b of the processing element 20b receives the decision log from the processing element 20a, and stores the decision log as the log 330b in the memory 33b.

  Then, the transaction execution unit 340a of the processing element 20a transmits a decision message requesting execution of the commit process (or abort process) for the transaction to all the processing elements 20c registered in the membership log (step S212).

  The transaction execution unit 340c of the processing element 20c receives this decision message, executes a transaction commit process (or abort process), and enters a pending state. In this pending state, the transaction execution unit 340c generates a log related to the decision, stores the log as the log 330c in the memory 33c using the synchronous neighbor-WAL protocol, and transmits the log to the processing element 20d (step S20). S213).

  The transaction execution unit 340d of the processing element 20d that has received this log stores the transmitted log as the log 330d in the memory 33d. Then, the transaction execution unit 340d executes all operations received from the processing element 20c and whose information is stored in the memory 33d as a log 330d.

  After execution of the operation, the transaction execution unit 340d transmits an ACK message indicating completion of the operation to the processing element 20c (step S214). After transmitting the ACK message, the transaction execution unit 340d deletes all logs related to this transaction from the memory 33d and releases the memory area.

  Upon receipt of this ACK message, the transaction execution unit 340c of the processing element 20c transmits a decision ACK message indicating that the decision processing has been completed in the processing element 20c and the processing element 20d to the processing element 20a (step S215). . After transmitting the decision ACK message, the transaction execution unit 340c deletes all logs related to this transaction from the memory 33c and releases the memory area.

  When the transaction execution unit 340a of the processing element 20a receives the decision ACK message from all the processing elements 20c registered in the membership log, the transaction execution unit 340a generates an end log indicating that the transaction decision processing has been completed, and the asynchronous neighbor-WAL. Using the protocol, the end log is stored in the memory 33a as the log 330a, and the end log is transmitted to the processing element 20b (step S216).

  The transaction execution unit 340b of the processing element 20b that has received the end log stores the transmitted end log in the memory 33b as the log 330b, and ends the transaction process using the BA-1.5 phase commit protocol.

  In step S213, the processing element 20c that is the primary cohort transmits a log related to the decision to the processing element 20d that is the backup cohort, and waits for an ACK message. However, the processing element 20d has some trouble, and the ACK When the message cannot be received, the decision ACK message is transmitted to the processing element 20a which is the primary master without deleting the log stored in the memory. This log is used later when the processing element 20d is restored.

  In this way, the processing element 20c that is the primary cohort goes through the pending state in the decision processing, but the processing result in the pending state of the processing element 20d that is the backup cohort does not affect the final result of the transaction.

  The name of the BA-1.5 phase commit protocol is named by counting the primary cohort decision process in such a pending state as 0.5 phase and the primary master decision process as 1 phase. is there.

  Further, in the transaction process, the log after the transaction process is completed is deleted from the memories 33c and 33d, so that the occurrence of memory overflow can be suppressed. If a memory overflow occurs, a part of the log may be written to the disk devices 40c and 40d.

  Next, recovery processing when a failure occurs in any of the primary master, the backup master, the primary cohort, and the backup cohort will be described. It is assumed that there are a plurality of processing elements 20c as primary cohorts and a plurality of processing elements 20d as backup cohorts.

(1) When a failure occurs in the backup master The failure of the processing element 20b, which is the backup master, has no effect on the transaction processing. The failure recovery processing unit 341a of the processing element 20a that is the primary master detects that a failure has occurred in the processing element 20b.

  The occurrence of the failure is determined by checking whether or not the socket communication for which the constant connection is established has been disconnected due to a timeout or the like, or whether or not a failure occurrence message indicating that a failure has occurred is received from the processing element 20b. Detected.

  If a failure is detected, the transaction execution unit 340a of the processing element 20a does not write a log to the memory 33b of the processing element 20b by the asynchronous neighbor-WAL protocol, but only to the memory 33a of the processing element 20a. Write the log, then continue normal processing.

(2) When a failure occurs in the primary master (2-1) When a failure occurs before the decision of the transaction FIG. 8 is a sequence diagram for explaining the return processing when a failure is detected in the primary master before the decision It is. As illustrated in FIG. 8, the failure recovery processing unit 341b of the processing element 20b that is the backup master detects a failure that has occurred in the processing element 20a that is the primary master (step S301).

  Then, the failure recovery processing unit 341b transmits an abort message for canceling the execution of the transaction to all the processing elements 20c that are the primary cohort registered in the membership log and the processing element 20d that is the backup cohort (step S302, step S302). S303). The failure recovery processing unit 341c of the processing element 20c and the failure recovery processing unit 341d of the processing element 20d receive this abort message and stop executing the transaction.

(2-2) When a failure occurs after deciding to execute a decision process and before receiving ACK messages from all primary cohorts Figure 9 shows receiving ACK messages from all primary cohorts after deciding to execute a decision process It is a sequence diagram explaining a return process when a failure is detected in the primary master before performing.

  Here, the processing element 20b that is the backup master receives the decision log from the processing element 20a after the processing element 20a that is the primary master has decided to execute the decision process, and therefore can detect the decision to execute the decision process. it can.

  As illustrated in FIG. 9, the failure recovery processing unit 341b of the processing element 20b that is the backup master detects a failure that has occurred in the processing element 20a that is the primary master (step S401). Then, the failure recovery processing unit 341b transmits a decision message to all the processing elements 20c registered in the membership log (step S402).

  Here, since the membership log is already stored in the memory 33b of the processing element 20b, the failure recovery processing unit 341b can detect all the processing elements 20c to which the decision message should be transmitted.

  The transaction execution unit 340c of the processing element 20c that has received this decision message executes the decision process, generates a log related to the decision process, and logs to the processing element 20d that is a backup cohort using the synchronous neighbor-WAL protocol. Is transmitted (step S403).

  The transaction execution unit 340d of the processing element 20d that has received this log executes all operations whose information is included in the log received from the processing element 20c, and transmits an ACK message indicating the completion of the operation to the processing element 20c ( Step S404).

  Upon receipt of this ACK message, the transaction execution unit 340c of the processing element 20c transmits a decision ACK message indicating that the decision processing has been completed in the processing element 20c and the processing element 20d to the processing element 20b (step S405). .

  Upon receiving the decision ACK message, the failure recovery processing unit 341b of the processing element 20b generates an end log indicating that the decision process is completed, stores the end log in the memory 33b as the log 330b, and ends the transaction process.

(3) When a failure occurs in the backup cohort In transaction processing, a failure in the backup cohort does not affect the primary master and the backup master. In fact, the primary master cannot detect the failure of the backup cohort because the synchronization process is not performed between the primary cohort and the backup cohort before the decision process is executed.

  FIG. 10 is a sequence diagram illustrating a return process when an abnormality is detected in the backup cohort. As shown in FIG. 10, first, the failure recovery processing unit 341c of the processing element 20c that is the primary cohort detects a failure that has occurred in the processing element 20d that is the backup cohort (step S501).

  When the failure recovery processing unit 341c detects a failure, the failure recovery processing unit 341c transmits an ACK message to the processing element 20a that is the primary master (step S502). Here, the transaction execution unit 340c retains the log stored in the memory 33c without erasing it.

  Thereafter, the failure recovery processing unit 341c detects the recovery of the processing element 20d (step S503). Then, after detecting the recovery, the failure recovery processing unit 341c transmits the log held in the memory 33c to the processing element 20d (step S504). The transaction execution unit 340d of the processing element 20d receives this log and stores it in the memory 33d, and updates the data to the latest state using the log.

(4) When a failure occurs in the primary cohort (4-1) When a failure occurs before the prepared state of the last operation The failure recovery processing unit 341a of the processing element 20a, which is the primary master, prepares the prepared state of the last operation. When a failure is detected in one of the processing elements 20c that is the primary cohort before (before log write (log n) in FIG. 7), the failure recovery processing unit 341a detects all of the processing elements 20c other than the processing element 20c that has detected the failure. An abort message is sent to the primary cohort and backup cohort requesting transaction abort.

(4-2) When a Failure Occurs Since the Prepared State of the Last Operation In this situation, the processing element 20a that is the primary master sends an ACK message for an operation execution request from all the processing elements 20c that are the primary cohort. Collect and send decision messages to all processing elements 20c.

  Since the execution of the operation is completed in each processing element 20c, if a failure occurs at this stage, the processing element 20a does not abort the transaction but continues it.

  FIG. 11 is a sequence diagram for explaining return processing when a failure is detected in the primary cohort after the prepared state of the last operation. As shown in FIG. 11, first, the failure recovery processing unit 341a of the processing element 20a that is the primary master detects a failure that has occurred in the processing element 20c that is the primary cohort (step S601).

  Then, the failure recovery processing unit 341a transmits a decision message requesting execution of the decision process for the transaction to the processing element 20d that is the backup cohort (step S602).

  The transaction execution unit 340d of the processing element 20d receives this decision message, executes a decision process, and transmits a decision ACK message to the processing element 20a (step S603). Here, the transaction execution unit 340d retains the log stored in the memory 33d without erasing it.

  The failure recovery processing unit 341c of the processing element 20c makes a log inquiry to the processing element 20d after recovering from the failure (step S604).

  When the failure recovery processing unit 341d of the processing element 20d receives this inquiry, it transmits a log to the processing element 20c (step S605). The failure recovery processing unit 341c of the processing element 20c uses this log to update the data to the latest state.

  The various processes described in the present embodiment can be realized by executing a program prepared in advance on a computer. In the following, an example of a computer that executes a program that implements the various processes will be described.

  FIG. 12 is a diagram illustrating a hardware configuration of a computer serving as the controllers 30a to 30d illustrated in FIG. This computer includes a network interface 100 that exchanges data with other computers via a network, a disk interface 101 that exchanges data with a disk device, a CPU (Central Processing Unit) 102, and a ROM (Read Only memory (RAM) 103 and RAM (Random Access Memory) 104 are connected by a bus 105.

  The ROM 103 stores a data storage control program 103a that exhibits the same functions as the functions of the controllers 30a to 30d. The data storage control program 103a may be stored in a distributed manner as appropriate.

  Then, when the CPU 102 reads out the data storage control program 103a from the ROM 103 and executes it, the data storage control process 102a is started and functions. Although the CPU 102 executes the data storage control program 103a here, an MCU (Micro Controller Unit) or an MPU (Micro Processing Unit) may execute the data storage control program 103a.

  The data storage control program 103a corresponds to the transaction execution unit 340a (340b, 340c, 340d) and the failure recovery processing unit 341a (341b, 341c, 341d) shown in FIG.

  Further, the CPU 102 stores the log 104a, the backup disk management data 104b, and the ACK management data 104c in the RAM 104, and reads out the data from the RAM 104 and executes various data processing.

  The log 104a, the backup disk management data 104b, and the ACK management data 104c are the log 330a (330b, 330c, 330d), the backup disk management data 331a (331b, 331c, 331d), and the ACK management data 332a (332b, 332b, 332d) illustrated in FIG. 332c, 332d).

  Next, an overhead comparison between the BA-1.5 phase commit protocol (BA-1.5PC), the two-phase commit protocol (2PC), and the early prepare protocol (EP) according to the present embodiment will be described. The overhead here includes both transaction operation processing, message exchange in commit processing, and forced disk writing.

  FIG. 13 is a diagram illustrating the overhead of committed transactions for each protocol. Here, P is the number of operations performed in the transaction, and N is the number of cohorts that perform the operation.

  As shown in FIG. 13, the numbers of messages transmitted in the BA-1.5 phase commit protocol, the two phase commit protocol, and the early prepare protocol are 3 + 3P + 4N, 2P + 8N, and 4P + 2N, respectively.

  The number of forced log writes to the disk device executed in the BA-1.5 phase commit protocol, the two phase commit protocol, and the early prepare protocol is 0, 4N + 1, and 2P + 2, respectively.

  The number of messages in the BA-1.5 phase commit protocol can be obtained from FIG. Further, in the BA-1.5 phase commit protocol, the asynchronous neighbor-WAL protocol is used, so that the forced log write to the disk device is unnecessary, and the number of forced log writes becomes zero.

  In the two-phase commit protocol, the cohort needs to execute a forced disk write twice during commit processing (once at voting and once at decision). On the other hand, the master needs to execute a forced disk write once during the decision.

  The master also needs to exchange two round-trip messages with each cohort. Therefore, when performing P operations in a transaction, the number of exchanged messages is 4N + P, and the number of forced disk writes is 2N + 1.

  When this two-phase commit protocol is applied to an autonomous disk, the primary cohort sends a message to the backup cohort and receives a response during operation and commit processing. In addition, a forced log write occurs twice during the commit process. Therefore, the number of messages in the two-phase commit protocol is 8N + 2P, and the number of forced log writes is 4N + 1.

  In the early prepare protocol, the cohort forcibly writes the log to the disk device together with the update data for each operation. The cohort then sends an ACK message for the operation to the master.

  At commit time, the cohort performs forced log write only for the aborted transaction and transmits an ACK message only for the aborted transaction. The forced log write performed by the master is a forced log write of the membership log at the start of the transaction and a forced log write of the decision log at the commit time. Therefore, the number of messages in the early prepare protocol is 2P + N, and the number of forced log writes is P + 2.

  When this early prepare protocol is applied to an autonomous disk, the primary cohort needs to send operation and decision messages to the backup cohort, so the number of messages is 4P + 2N, and the number of forced log writes is 2P + 2.

  As can be seen from FIG. 13, forced log write does not occur in the BA-1.5 phase commit protocol, and data processing can be performed at a higher speed than the two-phase commit protocol and the early prepare protocol. When N> 3 and P satisfies the relationship 2N + 3 <P <4N−3, the number of messages is minimized, and data processing can be performed more efficiently than the other two protocols.

  Next, a comparison of throughput between the BA-1.5 phase commit protocol, the two-phase commit protocol, and the early prepare protocol will be described. FIG. 14 is a diagram showing a comparison result of the throughput of each protocol. FIG. 14 shows throughputs related to insert processing of data of various sizes in an autonomous disk having 24 disk devices.

  As shown in FIG. 14, the throughput of the BA-1.5 phase commit protocol exceeds the throughputs of the two-phase commit protocol and the early prepare protocol for all data sizes. This is because the BA-1.5 phase commit protocol eliminates the complexity of message exchange in the commit process and eliminates the need for forced log writing.

  FIGS. 15, 16 and 17 are diagrams illustrating the throughput of the BA-1.5 phase commit protocol, the 2PC protocol, and the EP protocol, respectively. As shown in FIGS. 15, 16 and 17, it can be seen that in any protocol, the throughput increases as the number of disk devices (nodes) increases.

  The throughput in the BA-1.5 phase commit protocol is larger than the throughput of the 2PC protocol and the EP protocol regardless of the number of disk devices. In this manner, data processing can be performed at high speed by applying the BA-1.5 phase commit protocol to the autonomous disk.

  As described above, in this embodiment, the processing element 20a that is a primary master transmits an operation execution request message for data, and one or more processing elements 20b that are primary cohorts receive an operation execution request message. When the operation is received, the operation is executed, and an ACK message for the operation execution request message is transmitted to the processing element 20a every time the execution of the operation is completed, and log information related to the operation is generated to include the log information. And one or more processing elements 20d that are backup cohorts receive messages sent by one or more processing elements 20c. Log information relating to each operation is stored in the memory 33d, and the processing element 20a receives the ACK message transmitted by the one or more processing elements 20c and determines whether or not the execution of all the operations is completed; After execution of all the operations is completed, a decision process execution request message for the operation is transmitted. When one or more processing elements 20c receive the decision process execution request message transmitted by the processing element 20a, the decision process is performed. And updating the data stored in the disk device 40c connected to each of the one or more processing elements 20c, and generating log information related to the decision processing to generate the log The one or more processing elements 20d receive the message transmitted by the one or more processing elements 20c, and execute the decision process based on the log information stored in the memory 33d. Since the backup data stored in the disk device 40d connected to each of the one or more processing elements 20d is updated, an increase in overhead associated with the transaction is suppressed and data processing efficiency is improved. be able to.

  Further, in this embodiment, one or more processing elements 20d that are backup cohorts execute a decision process, and then transmit a decision completion message indicating the completion of the decision process, thereby one or more processing elements that are the primary cohort. When 20c receives a decision completion message from one or more processing elements 20d, it sends a synchronization completion message indicating that the data synchronization processing has been completed to the processing element 20a as the primary master. Thus, synchronization can be ensured between the data and the backup data.

  In this embodiment, the processing element 20a as the primary master generates membership log information indicating one or more processing elements 20c for transmitting the operation execution request message, and transmits the membership log information to the processing element 20b as the backup master. Then, the processing element 20b stores the membership log information in the memory 33b, and when the processing element 20a determines execution of the decision processing, it generates decision log information indicating the execution determination of the decision processing, and sends it to the processing element 20b. And when the processing element 20b detects a failure related to the processing element 20a after receiving the decision log information, the membership log information Since it was decided to send the execution request message Decision process to one or more processing elements 20c for operations based on, even if a failure of the processing elements 20a occurs, it is possible to quickly execute the decision process.

  Further, in this embodiment, after one or more processing elements 20c that are primary cohorts receive a decision processing execution request message from the processing element 20a that is primary master, one or more processing elements 20d that are backup cohorts. When a failure is detected, a decision completion message is transmitted to the processing element 20a. When the failure is recovered, one or more processing elements 20d are determined based on the log information stored in the memory 33d. Since the process is executed to update the backup data stored in the disk device 40d connected to each of the one or more processing elements 20d, the processing element 20d Even if according failure can be quickly execute the decision process.

  Further, in the present embodiment, after the processing element 20a that is the primary master receives all the response messages transmitted by the one or more processing elements 20c that are the primary cohort, one of the one or more processing elements 20c When a failure related to any of the above is detected, a decision processing execution message for the operation is transmitted to one or more processing elements 20d, and the one or more processing elements 20d transmit the decision processing transmitted by the processing element 20a. When the execution request message is received, the decision process is executed, and when the failure is recovered, the log information related to the decision process is transmitted to the processing element 20c where the failure has occurred, Since the processing element 20c receives the log information related to the decision processing, and executes the decision processing based on the received log information, the decision processing can be quickly performed even when a failure related to the processing element 20c occurs. Can be executed.

  Although the embodiments of the present invention have been described so far, the present invention may be implemented in various different embodiments in addition to the above-described embodiments within the scope of the technical idea described in the claims. It ’s good.

  For example, in step S215 of FIG. 7, the primary cohort sends a decision ACK message to the primary master after synchronizing data with the backup cohort, but waits for data synchronization with the backup cohort. Instead, a decision ACK message may be transmitted. Thereby, the time required for the decision processing can be further shortened.

  In addition, among the processes described in this embodiment, all or part of the processes described as being performed automatically can be performed manually, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method.

  In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.

  Each component of each illustrated device is functionally conceptual and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  Furthermore, each processing function performed in each device is realized by a CPU (or MCU or MPU) and a program that is analyzed and executed by the CPU (or MCU or MPU), Alternatively, it can be realized as hardware by wired logic.

(Supplementary Note 1) A data storage control device that performs control to store data and backup data of the data in different data storage units,
A first storage control unit for transmitting a plurality of operation execution request messages for data;
When the operation execution request message is received, the operation is executed, and a response message to the operation execution request message is transmitted to the first storage control unit every time the execution of the operation is completed. One or more second storage control units for generating information and transmitting a message including the log information;
One or more third storage control units that receive messages transmitted by the one or more second storage control units and store log information related to each operation in a memory;
With
The first storage control unit receives the response message transmitted by the one or more second storage control units, determines whether or not execution of all operations is completed, and executes all operations. After completion, send a decision processing execution message for the operation,
The one or more second storage control units execute the decision processing when receiving the decision processing execution request message transmitted by the first storage control unit, and the one or more second storage control units While controlling the update of the data stored in the data storage unit connected to each of the control unit, generate log information related to the decision processing and send a message including the log information,
The one or more third storage control units receive a message transmitted from the one or more second storage control units and execute a decision process based on log information stored in the memory. A data storage control device that performs update control of backup data stored in a data storage unit connected to each of the one or more third storage control units.

(Supplementary Note 2) After the execution of the decision process, the one or more third storage control units transmit a decision completion message indicating the completion of the decision process, and the one or more second storage control units When a decision completion message is received from the one or more third storage control units, a synchronization completion message indicating that data synchronization processing has been completed is transmitted to the first storage control unit. The data storage control device according to appendix 1, which is characterized by the above.

(Additional remark 3) It further has a 4th storage control part which memorizes membership log information which shows one or more 2nd storage control parts which transmit an execution request message of operation in a memory, The 1st storage control part Generates the membership log information and transmits it to the fourth storage control unit, and generates decision log information indicating the decision to execute the decision process when the decision to execute the decision process is generated, and stores the fourth log. When the fourth storage control unit detects a failure related to the first storage control unit after receiving the decision log information, the fourth storage control unit responds to the operation based on the membership log information. The decision process execution request message is transmitted to the one or more second storage control units. Over data storage control device.

(Supplementary Note 4) After the one or more second storage control units receive the decision processing execution request message from the first storage control unit, the faults related to the one or more third storage control units When the failure is detected, the decision completion message is transmitted to the first storage control unit. When the failure is recovered, the one or more third storage control units are stored in the memory. The decision processing is executed based on log information, and update control of backup data stored in a data storage unit connected to each of the one or more third storage control units is performed. 2. The data storage control device according to 2 or 3.

(Supplementary Note 5) The first storage control unit receives all response messages transmitted by the one or more second storage control units, and then includes the one or more second storage control units. When a failure related to any of the above is detected, a decision processing execution request message for an operation is transmitted to the one or more third storage control units, and the one or more third storage control units A second storage control in which the decision process is executed when the decision process execution request message transmitted by the first storage control unit is received, and the log information related to the decision process is restored from the failure when the failure is recovered; The second storage control unit that is transmitted to the unit and recovered from the failure receives the log information related to the decision process, and executes the decision process based on the received log information Data storage control apparatus according to any one of Appendices 1 to 4, characterized.

(Appendix 6) A data storage device that includes a plurality of data storage units and stores data and backup data of the data in different data storage units,
A first storage control unit for transmitting a plurality of operation execution request messages for data;
When the operation execution request message is received, the operation is executed, and a response message to the operation execution request message is transmitted to the first storage control unit every time the execution of the operation is completed. One or more second storage control units for generating information and transmitting a message including the log information;
One or more third storage control units that receive messages transmitted by the one or more second storage control units and store log information related to each operation in a memory;
With
The first storage control unit receives the response message transmitted by the one or more second storage control units, determines whether or not execution of all operations is completed, and executes all operations. After completion, send a decision processing execution message for the operation,
The one or more second storage control units execute the decision processing when receiving the decision processing execution request message transmitted by the first storage control unit, and the one or more second storage control units Update the data stored in the data storage unit connected to each of the control unit, generate log information related to the decision processing and send a message including the log information,
The one or more third storage control units receive a message transmitted from the one or more second storage control units and execute a decision process based on log information stored in the memory. A data storage device that updates backup data stored in a data storage unit connected to each of the one or more third storage control units.

(Supplementary Note 7) The one or more third storage control units, after executing the decision processing, transmit a decision completion message indicating completion of the decision processing, and the one or more second storage control units When a decision completion message is received from the one or more third storage control units, a synchronization completion message indicating that data synchronization processing has been completed is transmitted to the first storage control unit. The data storage device according to appendix 6, which is characterized.

(Supplementary Note 8) The first storage control unit further includes a fourth storage control unit that stores, in a memory, membership log information indicating one or more second storage control units that transmit the operation execution request message. Generates the membership log information and transmits it to the fourth storage control unit, and generates decision log information indicating the decision to execute the decision process when the decision to execute the decision process is generated, and stores the fourth log. When the fourth storage control unit detects a failure related to the first storage control unit after receiving the decision log information, the fourth storage control unit responds to the operation based on the membership log information. 8. The decision process execution request message is transmitted to the one or more second storage control units. Over data storage device.

(Supplementary note 9) The one or more second storage control units receive a decision processing execution request message from the first storage control unit, and then have a failure related to the one or more third storage control units. When the failure is detected, the decision completion message is transmitted to the first storage control unit. When the failure is recovered, the one or more third storage control units are stored in the memory. Appendices 6 and 7, wherein a decision process is executed based on log information to update backup data stored in a data storage unit connected to each of the one or more third storage control units Or the data storage device according to 8;

(Supplementary Note 10) The first storage control unit receives all response messages transmitted by the one or more second storage control units, and then includes the one or more second storage control units. When a failure related to any of the above is detected, a decision processing execution request message for an operation is transmitted to the one or more third storage control units, and the one or more third storage control units A second storage control in which the decision process is executed when the decision process execution request message transmitted by the first storage control unit is received, and the log information related to the decision process is restored from the failure when the failure is recovered; The second storage control unit that has been transmitted to the unit and recovered from the failure receives the log information related to the decision process, and executes the decision process based on the received log information. Data storage device according to any one of Appendices 6-9 characterized by.

(Supplementary note 11) A data storage method for storing data and backup data of the data in different data storage units,
An operation execution request message transmission step in which the first storage control unit transmits an operation execution request message for data;
When one or more second storage control units receive the operation execution request message, the one or more second storage control units execute the operation, and each time the execution of the operation is completed, a first response message to the operation execution request message is sent. To the storage control unit, generates log information related to the operation, transmits a message including the log information, and one or more third storage control units transmit the one or more second storage controls. An operation execution step of receiving a message transmitted by the unit and storing log information related to each operation in a memory;
The first storage control unit receives the response message transmitted by the one or more second storage control units, determines whether or not all the operations have been executed, and executes all the operations. A decision execution request message transmission step for transmitting a decision processing execution request message for the operation after completion;
When the one or more second storage control units receive the decision processing execution request message transmitted by the first storage control unit, the decision processing is executed, and the one or more second storage control units Updating the data stored in the data storage unit connected to each of the control units, generating log information related to the decision process, and transmitting a message including the log information; The storage control unit receives the message transmitted by the one or more second storage control units, executes a decision process based on the log information stored in the memory, and the one or more third storage units A decision execution step for updating backup data stored in a data storage unit connected to each of the storage control units;
A data storage method comprising:

(Supplementary Note 12) A decision completion message transmission step of transmitting a decision completion message indicating completion of the decision processing after the one or more third storage control units have executed the decision processing, and the one or more first storage control portions. When the two storage control units receive the decision completion message from the one or more third storage control units, the synchronization completion indicating that the data synchronization processing is completed to the first storage control unit The data storage method according to claim 11, further comprising a synchronization completion message transmission step of transmitting a message.

(Supplementary Note 13) In the operation execution request message transmission step, the first storage control unit generates membership log information indicating one or more second storage control units that transmit the operation execution request message. The fourth storage control unit stores the membership log information in a memory, and the decision execution request message transmission step is performed by the first storage control unit in the decision process. Decision log information indicating decision of execution of decision processing is generated and transmitted to the fourth storage control unit, and after the fourth storage control unit receives the decision log information, When a failure related to one storage control unit is detected, the decision processing for the operation is executed based on the membership log information. Data storage method according to appendix 11 or 12 and transmits the request message the one or more in the second storage control unit.

(Supplementary Note 14) In the decision execution step, the one or more second storage control units receive the decision processing execution request message from the first storage control unit, and then the one or more third storage control units When a failure related to the storage control unit is detected, the decision completion message is transmitted to the first storage control unit, and when the failure is recovered, the one or more third storage control units are: A decision process is executed based on the log information stored in the memory to update backup data stored in a data storage unit connected to each of the one or more third storage control units. The data storage method according to Supplementary Note 11, 12 or 13.

(Supplementary Note 15) In the decision execution request message transmission step, after the first storage control unit has received all response messages transmitted by the one or more second storage control units, the one or more ones When a failure related to any one of the second storage control units is detected, a decision processing execution request message for an operation is transmitted to the one or more third storage control units, and the decision execution step includes: When the one or more third storage control units receive the decision processing execution request message transmitted by the first storage control unit, the decision processing is executed, and the decision is made when the failure is recovered. The log information related to the process is transmitted to the second storage control unit recovered from the failure, and the second storage control unit recovered from the failure is involved in the decision processing. Data storage method according to the received log information, any one of Appendices 11 to 14, characterized in that to execute the decision processing based on the received log information.

(Supplementary Note 16) A data storage control program for performing control for storing data and backup data of the data in different data storage units,
An operation execution request message transmission procedure in which the first storage control unit transmits an operation execution request message for data;
When one or more second storage control units receive the operation execution request message, the one or more second storage control units execute the operation, and each time the execution of the operation is completed, a first response message to the operation execution request message is sent. To the storage control unit, generates log information related to the operation, transmits a message including the log information, and one or more third storage control units transmit the one or more second storage controls. An operation execution procedure for receiving a message transmitted by the unit and storing log information related to each operation in a memory;
The first storage control unit receives the response message transmitted by the one or more second storage control units, determines whether or not all the operations have been executed, and executes all the operations. After completion, a decision execution request message sending procedure for sending a decision processing execution request message for the operation;
When the one or more second storage control units receive the decision processing execution request message transmitted by the first storage control unit, the decision processing is executed, and the one or more second storage control units Updating the data stored in the data storage unit connected to each of the control units, generating log information related to the decision process, and transmitting a message including the log information; The storage control unit receives the message transmitted by the one or more second storage control units, executes a decision process based on the log information stored in the memory, and the one or more third storage units A decision execution procedure for updating backup data stored in a data storage unit connected to each of the storage control units;
A data storage control program for causing a computer to execute.

  As described above, the data storage control device and the data storage device according to the present invention are useful for a data storage system that can suppress an increase in overhead related to a transaction and improve the efficiency of data processing.

It is a figure which shows the function structure of the autonomous disk system which concerns on a present Example. It is explanatory drawing explaining Fat-Btree. It is a figure which shows the function structure of the processing elements 20a-20d shown in FIG. FIG. 4 is a diagram illustrating an example of backup disk management data 331a to 331d illustrated in FIG. 3. It is a figure which shows an example of the ACK management data 332a-332d shown in FIG. It is a sequence diagram explaining an asynchronous neighbor-WAL protocol. It is a sequence diagram explaining the BA-1.5 phase commit protocol which concerns on a present Example. It is a sequence diagram explaining a return process when a failure is detected in the primary master before the decision. FIG. 10 is a sequence diagram for explaining a return process when a failure is detected in the primary master before ACK messages are received from all primary cohorts after decision processing execution is determined. It is a sequence diagram explaining a return process when a failure is detected in a backup cohort. It is a sequence diagram explaining a return process when a failure is detected in the primary cohort after the prepared state of the last operation. It is a figure which shows the hardware constitutions of the computer used as the controllers 30a-30d shown in FIG. FIG. 4 is a diagram illustrating the overhead of committed transactions for each protocol. It is a figure which shows the comparison result of the throughput of each protocol. It is a figure which shows the throughput of BA-1.5 phase commit protocol. It is a figure which shows the throughput of 2PC protocol. It is a figure which shows the throughput of EP protocol.

Explanation of symbols

10a to 10c Client 100 Network interface 101 Disk interface 102 CPU
103 ROM
104 RAM
105 Bus 20a-20d Processing element 30a-30d Controller 31a-31d Network interface unit 32a-32d Disk interface unit 33a-33d Memory 330a-330d, 104a Log 331a-331d, 104b Backup disk management data 332a-332d, 104c ACK management data 34a to 34d Control unit 340a to 340d Transaction execution unit 341a to 341d Failure recovery processing unit 40a to 40d Disk device 400a to 400d Directory data 401a to 401d Directory backup data 402a to 402d Primary data 403a to 403d Backup data 50 Network

Claims (10)

A data storage control device that performs control to store data and backup data of the data in different data storage units,
A first storage control unit for transmitting a plurality of operation execution request messages for data;
When the operation execution request message is received, the operation is executed, and a response message to the operation execution request message is transmitted to the first storage control unit every time the execution of the operation is completed. One or more second storage control units for generating information and transmitting a message including the log information;
One or more third storage control units that receive messages transmitted by the one or more second storage control units and store log information related to each operation in a memory;
With
The first storage control unit receives the response message transmitted by the one or more second storage control units, determines whether or not execution of all operations is completed, and executes all operations. After completion, send a decision processing execution message for the operation,
The one or more second storage control units execute the decision processing when receiving the decision processing execution request message transmitted by the first storage control unit, and the one or more second storage control units While controlling the update of the data stored in the data storage unit connected to each of the control unit, generate log information related to the decision processing and send a message including the log information,
The one or more third storage control units receive a message transmitted from the one or more second storage control units and execute a decision process based on log information stored in the memory. A data storage control device that performs update control of backup data stored in a data storage unit connected to each of the one or more third storage control units.   The one or more third storage control units, after executing the decision processing, transmit a decision completion message indicating completion of the decision processing, and the one or more second storage control units When a decision completion message is received from the third storage control unit, a synchronization completion message indicating that data synchronization processing has been completed is transmitted to the first storage control unit. Item 4. The data storage control device according to Item 1.   A fourth storage control unit that stores, in a memory, membership log information indicating one or more second storage control units that transmit an operation execution request message; and the first storage control unit includes the member Ship log information is generated and transmitted to the fourth storage control unit. When execution of the decision process is determined, decision log information indicating determination of execution of the decision process is generated and transmitted to the fourth storage control unit. When the fourth storage control unit detects a failure related to the first storage control unit after receiving the decision log information, the fourth storage control unit executes a decision process for the operation based on the membership log information. 3. The data recording according to claim 1, wherein a request message is transmitted to the one or more second storage control units. The control device.   The one or more second storage control units detect a failure related to the one or more third storage control units after receiving a decision processing execution request message from the first storage control unit The decision complete message is transmitted to the first storage control unit, and when the failure is recovered, the one or more third storage control units are based on the log information stored in the memory. A decision process is executed to update control of backup data stored in a data storage unit connected to each of the one or more third storage control units. 4. The data storage control device according to 3.   The first storage control unit receives all response messages transmitted by the one or more second storage control units, and then receives the response message from any one of the one or more second storage control units. When such a failure is detected, a decision process execution request message for an operation is transmitted to the one or more third storage control units, and the one or more third storage control units perform the first storage control. When the decision process execution request message transmitted by the unit is received, the decision process is executed, and when the failure is recovered, log information related to the decision process is transmitted to the second storage control unit recovered from the failure. The second storage control unit recovered from the failure receives log information related to the decision processing, and executes the decision processing based on the received log information. Data storage control apparatus according to any one of claims 1 to 4. A data storage device having a plurality of data storage units and storing data and backup data of the data in different data storage units,
A first storage control unit for transmitting a plurality of operation execution request messages for data;
When the operation execution request message is received, the operation is executed, and a response message to the operation execution request message is transmitted to the first storage control unit every time the execution of the operation is completed. One or more second storage control units for generating information and transmitting a message including the log information;
One or more third storage control units that receive messages transmitted by the one or more second storage control units and store log information related to each operation in a memory;
With
The first storage control unit receives the response message transmitted by the one or more second storage control units, determines whether or not execution of all operations is completed, and executes all operations. After completion, send a decision processing execution message for the operation,
The one or more second storage control units execute the decision processing when receiving the decision processing execution request message transmitted by the first storage control unit, and the one or more second storage control units Update the data stored in the data storage unit connected to each of the control unit, generate log information related to the decision processing and send a message including the log information,
The one or more third storage control units receive a message transmitted from the one or more second storage control units and execute a decision process based on log information stored in the memory. A data storage device that updates backup data stored in a data storage unit connected to each of the one or more third storage control units.   The one or more third storage control units, after executing the decision processing, transmit a decision completion message indicating completion of the decision processing, and the one or more second storage control units When a decision completion message is received from the third storage control unit, a synchronization completion message indicating that data synchronization processing has been completed is transmitted to the first storage control unit. Item 7. The data storage device according to Item 6.   A fourth storage control unit that stores, in a memory, membership log information indicating one or more second storage control units that transmit an operation execution request message; and the first storage control unit includes the member Ship log information is generated and transmitted to the fourth storage control unit. When execution of the decision process is determined, decision log information indicating determination of execution of the decision process is generated and transmitted to the fourth storage control unit. When the fourth storage control unit detects a failure related to the first storage control unit after receiving the decision log information, the fourth storage control unit executes a decision process for the operation based on the membership log information. 8. The data storage according to claim 6, wherein a request message is transmitted to the one or more second storage control units. Apparatus.   The one or more second storage control units detect a failure related to the one or more third storage control units after receiving a decision processing execution request message from the first storage control unit The decision complete message is transmitted to the first storage control unit, and when the failure is recovered, the one or more third storage control units are based on the log information stored in the memory. 9. The decision processing is executed to update backup data stored in a data storage unit connected to each of the one or more third storage control units. The data storage device described.   The first storage control unit receives all response messages transmitted by the one or more second storage control units, and then receives the response message from any one of the one or more second storage control units. When such a failure is detected, a decision process execution request message for an operation is transmitted to the one or more third storage control units, and the one or more third storage control units perform the first storage control. When the decision process execution request message transmitted by the unit is received, the decision process is executed, and when the failure is recovered, log information related to the decision process is transmitted to the second storage control unit recovered from the failure. The second storage control unit recovered from the failure receives log information related to the decision processing, and executes the decision processing based on the received log information. Data storage device according to any one of claims 6-9.

Images