JP2011053932A - Storage system, data storage method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve system performance and resource usage efficiency in a storage system. <P>SOLUTION: A storage node includes a data storage device, in which a plurality of data sets are stored, and a data management device having data management parts which correspond to the plurality of data sets, respectively. The data management parts are provided with: data management system execution parts which achieve a data arrangement-managing function, which searches data included in the corresponding data sets, and a data exclusion-controlling function, which retrieves stops access to the data included in the corresponding data sets, on the basis of any of a plurality of data management systems; and system-changing parts which change the data management systems in the data management system execution parts from a first data management systems to a second data management systems among a plurality of data management systems. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a storage system, a data storage method, and a program, and more particularly, to a storage system, a data storage method, and a program for distributing and storing data in a plurality of storage nodes.

  A large-capacity storage system is known in which a plurality of computers (storage nodes) including a storage device for storing data are connected to a network.

  In the storage system, distributed data arrangement management manages to which storage node data is stored and to which storage node data read / write access by the user computer is distributed. Data transmitted from a user computer that wants to store data is divided into a plurality of data chunks and stored in a plurality of storage nodes by a distributed data arrangement management function in the storage system.

  User computers connected to the same network and using data use the distributed data placement management function to identify the storage node where the data is placed from the identifier that represents the desired data, and obtain the data from the storage node To do.

  According to the distributed data arrangement management function described in Non-Patent Document 2, when storing a fixed-length data chunk (chunk) in a storage node (GFS (Google File System) chunk server), the distributed data arrangement management apparatus (GFS master). The file name and the chunk index are transmitted as identifiers, and the chunk position information representing the target storage node is received.

  As an implementation method of the distributed data placement management function, an out-of-band method in which a placement management function that is centrally controlled in a place where it can be used from a user computer, and a part of the placement management function are delegated to each user computer, the user computer A distributed table system that directly accesses a storage unit is known.

  According to the out-of-bound method described in Non-Patent Document 2, when storing a fixed-length data chunk (chunk) in a storage node (GFS chunk server), a distributed data arrangement in which user computers (GFS clients) exist in the network The storage device is determined by accessing the management device (GFS master).

  According to the distributed table method described in Non-Patent Document 3, when storing a data chunk (value) of an arbitrary length in a node serving as both a user computer and a storage node, an identifier is used using a predetermined algorithm (Chord) within the node. The storage node is determined from (key).

  Patent Document 1 describes a method for ensuring fault tolerance by not only distributing data but also making data redundant. Fault tolerance refers to maintaining the state in which data can be provided to the user as a whole system even when a predetermined number of storage devices fail and data cannot be read. Non-Patent Document 1 describes a plurality of methods (RAID) for data distribution and redundancy at the block storage level. In Non-Patent Document 2, a distributed data arrangement management device (GFS master) and a storage node (GFS chunk server) communicate with each other during data storage, and a fixed-length data chunk (chunk) is copied to a plurality of storage nodes. A redundancy method for arranging the above is described.

  The performance of the storage nodes that make up the storage system may not be uniform. In order to reduce the cost of configuring the system, the system may be configured from a plurality of storage nodes having different processing capabilities, storage capacities, and data transfer rates. Hereinafter, such a system configuration is referred to as a “hetero configuration”.

  In a system based on a hetero configuration, data relocation is performed according to the frequency of use of data that changes with time in order to make the best use of a small number of high-performance storage nodes. By allocating frequently used data to a high-performance node, system resources can be efficiently used.

  In addition, access to data from the user computer is exclusively controlled in order to realize the performance and transactionality required by the user. Non-Patent Document 5 introduces the concept of degree or separability level with respect to separability of transactionality, and an exclusive control method for obtaining results of a plurality of processes without being influenced by each other by specifying the order Is described.

  Here, the exclusive control method includes a distributed exclusive control method and a centralized exclusive control method.

  As a distributed exclusive control method, when the distributed data arrangement management function has a distributed tree structure, according to the method described in Patent Document 2, the data arrangement management function is provided by a plurality of additional data accesses of the user computer. It can be prevented from being damaged.

  According to the distributed exclusive control system described in Non-Patent Document 4, the consistency of each data is secured by the storage node storing the data, and a single data update required by versioning or the like can be realized. it can. Further, according to the method described in Non-Patent Document 6, a distributed lock service can be provided by involving a plurality of nodes.

  According to the centralized exclusive control system described in Non-Patent Document 7, a single exclusive control function is realized in a system called a cache and lock service. In Non-Patent Document 7, it is pointed out that, as an advantage of the centralized lock management, a double configuration of the lock manager is possible and the efficiency of the deadlock processing is good.

Japanese Patent Laid-Open No. 11-085604 JP 2008-046700 A

David A. Patterson, Garth Gibson, and Randy H. Katz, “A case for redundant arrays of inexpensive disks (RAID),” ACM SIGMOD Record, Vol. 17, No. 3, pp. 109-116 Sanjay Ghemawat, Howard Gobioff, Shun-Tak Leung, “The Google File System,” SOSP 2003. Frank Dabek, Emma Brunskill, M. Frans Kaashoek, David Karger, Robert Morris, Ion Stoica, and Hari Balakrishnan, “Building Peer-to-Peer Systems With Chord, a Distributed Lookup Service,” Proceedings of the 8th Workshop on Hot Topics in Operating Systems (HotOS-VIII), May 2001. Giuseppe DeCandia, Deniz Hastorun, Madan Jampani, Gunavardhan Kakulapati, Avinash Lakshman, Alex Pilchin, Swaminathan Sivasubramanian, Peter Vosshall and Werner Vogels, “Dynamo: Amazon's Highly Available Key-value Store,” in the Proceedings of the 21st ACM Symposium on Principles, Stevenson, WA, October 2007. Jim Gray, Andreas Reuters, “Transaction Processing Concepts and Techniques (Up / Down)”, Nikkei Business Publications, 2001. Mike Burrows, “The Chubby Lock Service for Loosely-Coupled Distributed Systems,” OSDI’06: Seventh Symposium on Operating System Design and Implementation, Seattle, WA, November, 2006. Yotaro Nakayama, “Design and Implementation of Shared Disk Cluster PG-CALS with PostgreSQL”, UNISYS TECHNOLOGY REVIEW, Nihon Unisys, No.94, NOV. 2007 Leslie Lamport, “The Part-Time Parliament,” ACM Transactions on Computer Systems, Vol. 16, No.2, 133-169, May 1998.

  The storage system collectively stores data accessed by each of a plurality of data utilization processes. Here, a set of data accessed by each data use process is referred to as a “data set”. That is, the storage system stores a plurality of data sets.

  The distributed exclusive control method has high throughput, but communication between distributed functions is large and latency is long. Therefore, it is inefficient for a small amount of access. On the other hand, the centralized exclusive control method can process a small amount of access at high speed, but the exclusive control function becomes a bottleneck for a large amount of access, and the performance of the entire system is lowered. In particular, when the same exclusive control method is applied to the entire system when storing a large number of data sets having different uses and access amounts in a single storage system, an exclusive control method that is not suitable for the application is introduced.

  In addition, distributed data arrangement management has high throughput. In addition, by providing a part of the data allocation management function to many computers in the system, it becomes easy to search the data allocation management function from within the system, and it is easy to change the computer that provides the data allocation management function. Become. However, according to the distributed data arrangement management, more computer resources are required as compared with the centralized type.

  On the other hand, according to the centralized data arrangement management, the resources used for data arrangement management can be accommodated in a small number of computers as compared with the distributed type. However, according to the centralized data arrangement management, the latency for each access is large and becomes a bottleneck, so the throughput is also low. Also, a single data location management function that exists in the system must be searched, and if the data location management function is moved to another computer, the user computer cannot find the data location management function. The system becomes impossible to use.

  In addition, according to the distributed exclusive control method or the data arrangement management method, it is necessary to operate many computers even when there is little access to any stored data set. Therefore, according to the distributed type, power consumption is increased when the same number of accesses is processed as compared with the centralized type.

  From the above, the storage system cannot provide an exclusive control method according to the usage tendency that changes with time for each data set, and there is a problem that the performance is low. In addition, the storage system cannot provide a data arrangement management method according to the usage tendency that changes with time for each data set, and there is a problem that the efficiency of resource utilization is low. Furthermore, the storage system has a problem in that it cannot perform computer power usage control according to the usage trend that changes with time for each data set, and power consumption is large.

  Therefore, in the storage system, it is a problem to improve system performance and resource utilization efficiency. An object of the present invention is to provide a storage system that solves this problem.

The storage node according to the first aspect of the present invention is:
A data storage device storing a plurality of data sets;
A data management device having a data management unit corresponding to each of the plurality of data sets;
The data management unit includes a data placement management function for searching for data included in a corresponding data set, and a data exclusive control function for temporarily stopping access to data included in the corresponding data set. A data management method execution unit realized based on one of the methods;
The data management method execution unit further includes a method change unit that changes the data management method from the first data management method of the plurality of data management methods to the second data management method.

  The storage system according to the second aspect of the present invention includes a plurality of storage nodes according to the first aspect connected to each other via a network.

In the data storage method according to the third aspect of the present invention, a data management unit corresponding to each of a plurality of data sets stored in a storage data storage device includes:
A data placement management step of searching for data contained in a corresponding data set based on one of a plurality of data management methods;
A data exclusive control step of temporarily stopping access to data included in a corresponding data set based on any of the plurality of data management methods;
Changing a data management method in the data arrangement management step and the data exclusive control step from a first data management method of the plurality of data management methods to a second data management method;

The program according to the fourth aspect of the present invention is:
For the CPU corresponding to each of the plurality of data sets stored in the data storage device of the storage,
A data placement management process for searching for data contained in a corresponding data set based on one of a plurality of data management methods;
A data exclusive control process for temporarily stopping access to data included in a corresponding data set based on one of the plurality of data management methods;
A process of changing a data management system in the data arrangement management process and the data exclusive control process from a first data management system of the plurality of data management systems to a second data management system is executed.

  According to the storage system, data storage method and program of the present invention, system performance and resource utilization efficiency in the storage system can be improved.

1 is a block diagram showing a configuration of a storage system according to a first embodiment. 2 is a block diagram showing a configuration of a storage node in the storage system according to the first embodiment of the present invention. FIG. FIG. 2 is a block diagram illustrating a configuration of a data management device in a storage node in the storage system according to the first embodiment of the present invention. It is a figure for demonstrating an example of the distributed data arrangement management function in the storage system which concerns on the 1st Embodiment of this invention. 3 is a block diagram showing a configuration of a method change unit in the storage system according to the first embodiment of the present invention. FIG. It is a sequence diagram which shows operation | movement of the system change part in the storage system which concerns on the 1st Embodiment of this invention. It is a block diagram which shows the structure of the system change part in the storage system which concerns on the 2nd Embodiment of this invention. It is a block diagram which shows the structure of the storage system which concerns on the 3rd Embodiment of this invention. It is a block diagram which shows the structure of the storage node in the storage system which concerns on the 4th Embodiment of this invention.

  The storage node in the first development form is preferably a storage node according to the first aspect.

The storage node in the second deployment form is
Each of the plurality of data sets includes data identified by a data identifier;
The data management device further comprises a data set identification unit that returns a data set identifier that identifies a data set to which the data belongs among the plurality of data sets when receiving access to data from a data utilization process,
The data management unit searches for a data storage device in which the data is stored with reference to data arrangement information indicating the correspondence between the data and the data storage device in which the data is stored and the data identifier, It is preferable to realize the data arrangement management function.

  In the storage node of the third development form, the data management unit refers to the exclusive control information for the stored data, and temporarily stops access to the data by the data use process, so that the data exclusive control function Is preferably realized.

The storage node in the fourth deployment form is
The data management unit further includes a method selection unit that generates a data management method identifier representing the data management method from the data management method information;
Preferably, the data management method execution unit realizes the data arrangement management function and the data exclusive control function based on a data management method specified by the data management method identifier of the plurality of data management methods. .

  The storage system according to the fifth deployment mode preferably includes a plurality of the storage nodes connected to each other via a network.

The storage system in the sixth deployment form is
Distributed data that realizes the data allocation management function and the data exclusive control function by exchanging information between the plurality of data management methods at least between the distributed data management functions of a plurality of storage nodes via a network. Management method,
A centralized data management method in which at least one storage node is stubbed by delegating a request for these functions to another storage node without realizing the data arrangement management function or the data exclusive control function And preferably.

The storage system of the seventh deployment form is
The method change unit transmits data arrangement information and exclusive control information in at least one storage node in which the data management unit operates in the distributed data management method to the data management unit of another storage node, and after transmission ends, A centralized change unit that stubs the data management unit by delegating the data placement management function to another storage node;
Among the data management units that operate in the centralized data management method, a part of the data allocation information or exclusive control information held by the data management unit that implements the data allocation management function or the data exclusive control function is converted into a stub. It is preferable to further include a decentralized change unit that transmits the data to the other data management unit and after the transmission ends, causes the other data management unit to realize the data arrangement management function or the data exclusive control function.

The storage system of the eighth deployment form is
The method change unit generates a static point information generation unit that generates the data location information of the transmission source storage node at a specific time and the value of the exclusive control information as static point information;
A quiesce point information transmission unit for transmitting quiesce point information to other storage nodes;
A quiesce point information receiving unit that receives and holds quiesce point information transmitted from other storage nodes;
A management difference information generating and holding unit that compares the value at the current time with access from the user computer and the static point information, and generates and holds the difference between the static point information and the value at the current time as management difference information;
A management difference information transmission unit for transmitting management difference information to another storage node;
A management difference information receiving unit for receiving management difference information transmitted from another storage node;
Based on the still point information at the first time and the management difference information from the first time to the second time, the data arrangement information and the exclusive control information at the second time are restored as new still point information. A management difference information application unit;
The quiesce point information generating unit generates quiesce point information, transmits the value of the quiesce point information to the transmission destination storage node, and changes to the data arrangement information being transmitted and the exclusive control information as the first management difference information, the transmission source storage When the management difference information exists in the management difference information generation holding unit in the transmission source storage after the transmission of the still point information value is held in the management difference information generation holding unit in the node, the transmission difference storage node stores the management difference information. Management difference information receiving unit, and the management difference information applying unit in the storage node of the transmission destination reproduces the management difference information at the time of transmission in the transmission destination storage based on the received management difference information and the received still point information. In operation, the size of the management difference information is equal to or smaller than the predetermined size, and the number of transmissions of the management difference information is equal to or larger than the predetermined number. After repeated until, change the data management method in the data management method execution unit preferably further includes a method stop switch resumption unit that transmits the remaining management difference information.

  In the storage system of the ninth expansion mode, it is preferable that the centralized change unit stubs the data arrangement management function in the method stop switching restart unit.

The storage system of the tenth deployment form is
The data set identification unit further includes an access waiting unit uniquely associated with each of the plurality of data sets;
The data use process accesses the access waiting unit corresponding to the data set to which the data to be used belongs,
The data set identification unit may specify a data set based on an access waiting unit that has received access.

  In the storage system of the eleventh development mode, the data set identification unit may specify a data set by receiving data set identification information from the data utilization process together with access.

  In a storage system according to a twelfth development mode, the data set identification unit holds data / data set correspondence information that can identify a data set from the data identifier, and includes a data identifier included in an access from a data use process The data set may be specified based on the data / data set correspondence information.

  In the storage system of the thirteenth development mode, the data set identification unit can identify the data set from the user identification information that identifies the data use process or the user computer on which the data use process operates. The data set may be specified based on the user identification information obtained by receiving access from the data use process and the client data set correspondence information.

The storage system of the 14th deployment form is
The plurality of storage nodes include a high-speed and low-capacity storage node having a high-speed and low-capacity data storage device, and a high-speed and low-capacity storage node having a low-speed and large-capacity data storage device,
When the data set distributed and stored in the high-speed and low-capacity storage is the distributed data management method, and the data set is moved to the low-speed and large-capacity storage, the data management unit corresponding to the data set is It is preferable to further include a management method change unit for the data management method.

  In a storage system according to a fifteenth development mode, the management method changing unit uses the centralized data management method for data sets distributed and stored in the low-speed large-capacity storage, and moves the data sets to the high-speed and low-capacity storage. In this case, it is preferable that the data management unit corresponding to the data set is the distributed data management method.

The storage system of the sixteenth deployment form is
The storage node executes a specific data management method of the plurality of data management methods with low power consumption, and a specific data management method execution unit;
A general data management method execution unit for executing a data management method other than the predetermined management method;
It is preferable to further include a power control unit that shifts the general data management method execution unit to a low power consumption state when a data management method other than the predetermined management method is not operating in the own node. .

  The data storage method according to the seventeenth development form is preferably the data storage method according to the third aspect.

  The program according to the eighteenth development form is preferably a program according to the fourth viewpoint.

(First embodiment)
A storage system according to a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the storage system according to this embodiment. Referring to FIG. 1, the storage system includes a plurality of storage nodes 70a to 70c.

  Each of the storage nodes 70a to 70c includes data management devices 10a to 10c and data storage devices 60a to 60c, respectively. FIG. 1 also shows a user computer 90 connected to the storage system via the network 80. The data utilization process 91 of the user computer 90 performs access to the storage system including a data identifier and an identifier representing a data operation, thereby storing data, obtaining the stored data, updating the stored data, The stored data is erased.

  FIG. 2 is a block diagram showing the configuration of the storage node in the storage system according to this embodiment. Each of the storage nodes 70a to 70c in FIG. 1 corresponds to the storage node 70 in FIG. Referring to FIG. 2, the storage node 70 includes a CPU 71, a memory 72, an I / O controller 73, a data storage device 60, and a communication controller 74. The storage node 70 is configured by a computer as an example.

  The storage node 70 transmits / receives data to / from the network 80 via the communication controller 74 and stores / reads data to / from the data storage device 60 via the I / O controller 73. Such control may be performed by driving the CPU 71 by software stored in the memory 72.

  The data storage devices 60a to 60c are, for example, hard disk drives, flash memories, DRAMs, MRAMs, FeRAMs, storage devices coupled to a RAID controller, physical media capable of recording data such as magnetic tapes, or external storage nodes It is a control apparatus which records data on the medium installed in. Note that the memory 72 and the storage device 60 may be realized by the same hardware.

  The network 80 can be realized by an upper level protocol such as Fiber Channel, FCoE (Fibre Channel over Ethernet (registered trademark)), InfiniBand, or TCP / IP on the Ethernet. However, the implementation method of the network 80 is not limited to these.

  Data recorded in the storage system is distributed and stored in the data storage devices 60a to 60c of the storage nodes 70a to 70c. Data is managed by a fixed-length or arbitrary-length data chunk and a data identifier that identifies the data chunk. As an example, the data chunk may be a file and the data identifier may be an i-node number. Alternatively, the data chunk may be a 4 kilobyte long block, and the data identifier may be a volume name and block number pair.

  All data belongs to a single data set. One data set can include a plurality of data. The data set may be data in a virtual volume, for example. The data set may be a data set managed by one application software or middleware software of the database management system. Furthermore, when a single resource is shared by a plurality of companies, data belonging to each company may be used as a data set. However, the data set is not limited to these.

  FIG. 3 is a block diagram showing the configuration of the data management apparatus 10 in the storage node 70 in the storage system according to this embodiment. Each of the data management devices 10a to 10c in FIG. 1 corresponds to the data management device 10 in FIG. The data management apparatus 10 in FIG. 1 may be realized using the CPU 71 in FIG. Referring to FIG. 3, the data management device 10 includes a data set identification unit 51 and data management units 50a to 50c. Each of the data management units 50a to 50c corresponds to a stored data set. The data management units 50a to 50c are collectively referred to as the data management unit 50.

  Referring to FIG. 3, the data management unit 50 includes a method selection unit 40, a data management method execution unit 20, and a method change unit 30. The data management method execution unit 20 has a data exclusive control function and a data arrangement management function.

  Based on the access from the data utilization process 91, the data set identification unit 51 identifies which data set the data is included in. For example, the data set identification unit 51 may include a plurality of access waiting units (not shown) that are uniquely associated with the data set. At this time, the data use process 91 accesses the access waiting unit corresponding to the data set to which the data to be used belongs, and the data set identifying unit 51 identifies the data set by the access waiting unit that has received the access. Good.

  When the network protocol is TCP / IP, for example, a TCP port number or a virtual IP address may be used for identification by the access waiting unit. At this time, the data set identification unit 51 can recognize that connections to different ports or addresses are accesses to different data sets.

  The data set identification unit 51 may identify the data set by receiving the data set identification information from the data use process 91 together with the access. As an example, the data set identification information can be generated from a user identifier using the process 91. When the network protocol is HTTP, for example, the user can be specified by using COOKIE.

  The data set identification unit 51 holds data / data set correspondence information (not shown) that can specify a data set from the data identifier, and also includes the data identifier and the data / data set included in the access from the data use process 91. The data set may be specified based on the correspondence information. For example, the first few bytes of the data identifier may represent a data set. Further, when the data set is a virtual volume and the network protocol is SCSI, the virtual volume number may be embedded during access.

  The data set identification unit 51 holds client data set correspondence information (not shown) that can identify a data set from the data identification process 91 or user identification information that identifies the user computer 90 on which the data utilization process 91 operates. At the same time, the data set may be specified based on the user identification information obtained by receiving access from the data use process 91 and the client / data set correspondence information. For example, the address of the user computer 90 or the network address can be used as the user identification information.

  Moreover, you may make it use the above-mentioned data set identification part 51 combining suitably. For example, both the address of the user computer 90 and the user ID may be used. Further, the above-described data set identification unit 51 may be used step by step. For example, when a data set is specified by a port address to wait for but the data indicated by the data identifier included in access is not included in the specified data set, the storage system uses the user computer address. The data set may be identified by.

  The data arrangement for the storage nodes 70a to 70c is determined by the data arrangement management function operating in the storage nodes 70a to 70c, and in particular, the correspondence between the data identifier and the storage node is managed by the function. The data arrangement management function is realized by one of a plurality of data arrangement management methods.

  As the data arrangement management method, for example, a distributed data arrangement management method or a centralized data arrangement management method can be used. The distributed data arrangement management method is realized by exchanging information between data arrangement management functions of a plurality of storage nodes via a network. On the other hand, the centralized data arrangement management method delegates a request for the function to the data arrangement management function in another storage node without realizing the data arrangement management function in at least one storage node (that is, the stubbed data management unit). Is the method.

  FIG. 4 is a diagram for explaining an example of the distributed data arrangement management function in the storage system according to the present embodiment. As an example, the data arrangement management function can be realized by a distributed hash table, which is a kind of distributed data management method. The data arrangement management function at this time will be described with reference to FIG.

  Here, it is assumed that the data arrangement management function for the data set ds1 is realized by a distributed hash table. The data arrangement management function based on the distributed hash table includes a hash function h (d) that outputs an integer value for the byte string d and a node allocation table based on the range of output values of the hash function. With reference to FIG. 4, the operation for determining the target node will be described.

  The user computer 90 transmits a search request including a data identifier to one storage node. Here, it is assumed that the storage node that receives the search request is the storage node 70a. In the storage node 70a, the data set identification unit 51 identifies to which data set the data indicated by the data identifier belongs. Next, the data arrangement management function for managing the data arrangement of the data set performs data arrangement from the node allocation table based on the hash value x output when the data identifier d is input to the hash function h (d). Determine the storage node. Here, it is assumed that the data arrangement management function of the storage node 70b manages the storage node allocation of the hash value x. At this time, the data arrangement management function operating in the storage node 70a delegates the search request to the data arrangement management function of the storage node 70b.

  The data arrangement management function of the storage node 70b manages an allocation table composed of a pair of a data identifier d and a storage node that stores data, and determines a storage node based on the allocation table. The data arrangement management function of the storage node 70b returns an identifier for identifying the storage node to the user computer 90. As described above, the data use process 91 of the user computer 90 can specify and use the data position.

  As an example, the data arrangement management function can be realized by a centralized table management system which is a kind of centralized data management system. According to the centralized table management method, the data arrangement management function of a single storage node (for example, storage node 70a) manages the data arrangement. When accessed, the data arrangement management functions of the other storage nodes send the information to the data arrangement management function of the storage node 70a and delegate the data arrangement management. In this way, a storage node that has been delegated the management function without performing any processing relating to data management by itself is referred to as “stubbing”.

  The data arrangement management function of the storage node 70a manages a table from which a node including the data can be extracted from the data identifier. By referring to the table, the correspondence between the data identifier and the storage node is managed. Manage. In addition, the data allocation management function of some storage nodes holds a copy of some or all of the data allocation management information in case the storage node 70a fails, and always confirms the life and death of the storage node 70a. You may make it do.

  Further, the above distributed data management method and the centralized data management method may be used in combination. For example, some storage nodes may perform distributed data management, and the data arrangement management functions of the remaining storage nodes may delegate processing to the nodes that perform distributed data management. The method of the data arrangement management function is not limited to the above method.

  The data utilization process 91 of the user computer 90 can perform data access while satisfying the transaction property by transmitting information on the separation property or consistency constraint of the transaction property for the access. This is realized by exclusive control. For example, the degree of influence of a plurality of processes is controlled by designating a separability level indicating when a change in data by one access becomes visible to other accesses. This is realized by a lock for each data or data arrangement management function.

  Exclusive control for the data of the storage nodes 70a to 70c is managed for each data set by a data exclusive control function (see FIG. 3). The data exclusive control function is realized by one of a plurality of data exclusive control methods.

  As the data exclusive control method, for example, a distributed exclusive control method or a centralized exclusive control method can be used. The distributed exclusive control method is realized by exchanging information between distributed exclusive control functions of a plurality of storage nodes via a network. On the other hand, the centralized distributed exclusive control method delegates a request for the function to a data arrangement management function in another storage node without realizing a data exclusive control function in at least one storage node (that is, a stubbed data management unit) Is the method.

  For example, the data exclusive control function can be realized by applying the distributed data arrangement management function. In this case, each storage node manages a lock table from which lock information for data stored in the storage node can be extracted. The lock information includes a value that identifies the data use process 91 that currently holds the lock for the data. By using the lock table, the data identifier and access permission are managed. When the storage system creates multiple copies of the same data and stores them in multiple storage nodes, one of the replicas is set as primary data in advance, and lock information is stored in the storage node that stores the primary data It shall be.

  Access from the user computer 90 first reaches the storage node that holds the data, that is, the storage node that holds lock information for the data, by the distributed data arrangement management function. The storage node scans the lock table and determines whether or not access to the access is possible.

  According to the distributed data exclusive control function described in Non-Patent Document 6, the data exclusive control function in a group composed of a plurality of storage nodes (for example, five) shares a copy of lock information for the data set. . One of the storage nodes of the group is selected as a master by the distributed agreement algorithm, and the selected master manages the lock table. A process involving a change in exclusive control access reaches the master node, and the lock information change is propagated by the master to the storage nodes in the group. Processing that does not involve a change can be processed using a copy of lock information at any node in the group.

  As an example, the data exclusive control function can be realized by a centralized lock manager which is a kind of centralized data management system. According to the centralized lock manager, the data exclusive control function of the only storage node (for example, the storage node 70a) manages the exclusive control of data. When accessed, the data exclusive control functions of the other storage nodes send the information to the data exclusive control function of the storage node 70a and delegate data exclusive control. In this way, the storage node in which the management function has been delegated without performing the processing related to the data exclusive control function is referred to as “stubbing” as in the case of data arrangement management.

  The data exclusive control function of the storage node 70a manages a lock table that can extract lock information for the data from the data identifier. The lock information includes a value for identifying a data using process that currently holds a lock for the data. By referring to the lock table, the data identifier and access permission are managed. In addition, when a certain access reaches the data exclusive control function and the lock for the access is already held from the lock table and the data access cannot be permitted, the information indicating the access is put in a queue for the data. When the data is registered and the lock on the data is released, the processing for the waiting process registered in the queue may be continued. In addition, the data exclusion control function of some storage nodes holds a copy of part or all of the data exclusion control function and always checks whether the storage node 70a is alive or not in case the storage node 70a fails. It may be.

  Further, the distributed data exclusive control method and the centralized data exclusive control method may be used in combination. For example, some storage nodes can perform distributed data exclusive control, and the data arrangement management functions of the remaining storage nodes can delegate (stubbing) processing to nodes that perform distributed data exclusive control. The method of the data exclusive control function is not limited to the above method.

  Hereinafter, the data arrangement management function and the data exclusive control function are collectively referred to as a data management function. Changing the method of the data management function means changing the data arrangement management function from the X method to the Y method and changing the data exclusive control function from the Z method to the W method. Changing the data management function from the distributed system to the centralized system means changing the data arrangement management function from the distributed system P to the centralized system Q and changing the data exclusive control function from the distributed system R to the centralized system S. On the other hand, changing the data management function from the centralized system to the distributed system means changing the data arrangement management function from the centralized system Q to the distributed system P and changing the data exclusive control function from the centralized system S to the distributed system R. Say.

  The method change unit 30 changes the method of the data management function used by the data management method execution unit 20. FIG. 5 is a block diagram showing a configuration of the method changing unit 30 in the storage system according to the present embodiment. Referring to FIG. 5, the method changing unit 30 includes a still point information generating unit 31, a management difference information generating and holding unit 32, a still point information transmitting unit 33, a still point information receiving unit 34, a management difference information transmitting unit 35, and a management difference. An information receiving unit 36, a management difference information applying unit 37, and a method stop switching restarting unit 38 are further included.

  The quiesce point information generation unit 31 generates the values of the data arrangement information 22 and the exclusive control information 24 of the transmission source storage node at a specific time as quiesce point information. The quiesce point information transmission unit 33 transmits the quiesce point information to another storage node.

  The quiesce point information receiving unit 34 receives and holds quiesce point information transmitted by other storage nodes.

  The management difference information generation / holding unit 32 compares the access from the user computer 90, the still point information, and the above value at the current time, and generates and holds the difference between the still point information and the above value at the current time as the management difference information. To do. The management difference information transmission unit 35 transmits the management difference information to another storage node.

  The management difference information receiving unit 36 receives the management difference information transmitted by other storage nodes. The management difference information application unit 37, based on the still point information at the first time (t1) and the management difference information from the time t1 to the second time (t2) thereafter, the data arrangement information 22 at the time t2. The exclusive control information 24 is restored as new still point information.

  The method changing unit 30 having such a configuration changes the data management method as follows. FIG. 6 is a sequence diagram showing the operation of the method changing unit 30 in the storage system according to the present embodiment. With reference to FIG. 6, the operation of the method changing unit 30 will be described.

  The method is changed by exchanging the data arrangement information 22 and the exclusive control information 24 (hereinafter referred to as “data management information”) and switching the algorithm thereafter.

  First, in order to realize the post-change method, it is determined which storage node should hold which data management information, and each storage node creates a management information transmission plan.

  Next, each storage node generates quiesce point information by the quiesce point information generation unit 31. Next, each storage node transmits a necessary one of the values included in the quiesce point information to the target transmission destination storage node based on the management information transmission plan.

  A change in the data management information due to the access from the user computer 90 to the data arrangement information 22 and the exclusive control information 24 being transmitted is held in the management difference information generation holding unit 32 of the transmission source storage node as the first management difference information. The data management information necessary for access from the user computer 90 at that time is first scanned and returned by the management difference information, and if no value is held in the management difference information, the nearest still point information Returns the value from.

  When the data management information is updated during the transmission of the quiesce point information after the transmission of the necessary value of the quiesce point information, the management difference information exists in the management difference information generation holding unit 32 of the transmission source storage.

  If the management difference information exists and the size is equal to or larger than the predetermined size, or if the management difference information has not been transferred a predetermined number of times, the transmission source storage node manages the transmission difference storage node. Send difference information. The change of the data management information by the access from the user computer 90 during the transmission of the management difference information is held as the second management difference information. Further, the data management information necessary for access from the user computer 90 at the time point is when the second management difference information is first scanned and returned, and the value is not held in the second management difference information. Scans the first management difference information and returns its value, and if it does not exist in the first management difference information, returns the value of the most recent still point information.

  In the storage node of the transmission destination, the management difference information receiving unit 36 receives the management difference information, and the management difference information applying unit 37 determines the management difference information at the time of transmission based on the received management difference information and the received still point information. Reproduce to the destination storage.

  After the transmission of the first management difference information, if the second management difference information exists as described above, the management difference information is equal to or smaller than a predetermined size and the management difference information is a predetermined number of times. Until the number of transmissions exceeds a predetermined number of times (No in step S11 in FIG. 5), the first management difference information is replaced with the second management difference information, and the second management difference information is transmitted and the third management difference information is transmitted. The operation of creating difference information is repeated. On the other hand, when the size of the management difference information is equal to or smaller than the predetermined size and the transmission of the management difference information is equal to or larger than the predetermined number of times (Yes in step S11 in FIG. 5), this is required. Stop operation.

  The method stop switching / resuming unit 38 changes the data management function. First, the method stop switching / resuming unit 38 stops data management. Next, new data management is performed using the received data management information. The method stop switching / resuming unit 38 performs synchronization by communicating with a data management function in another storage node if necessary. Next, new data management is resumed. Thereby, the system of the data arrangement management function and / or the data exclusive control function can be changed.

  The storage system according to the present embodiment can provide an exclusive control method according to a usage tendency that changes with time for each data set, and thus the performance of the storage system can be improved. In addition, according to the storage system according to the present embodiment, it is possible to provide a data arrangement management method according to a usage trend that changes with time for each data set, so that it is possible to improve resource utilization efficiency in the storage system. it can.

  The reason is that the data set identification unit 51 identifies the data set to be accessed, and a different data management method can be applied to each data set. Further, the method change unit 30 can dynamically change the data management method for each data set. Furthermore, since the system change unit 30 uses the quiesce point information to shorten the period during which the system stops access, the system change unit 30 can also prevent the system performance from being degraded due to the system change.

(Second Embodiment)
A storage system according to a second embodiment of the present invention will be described with reference to the drawings. FIG. 7 is a block diagram showing a configuration of the method changing unit 30 in the storage system according to the present embodiment. Referring to FIG. 7, the method change unit 30 further includes a decentralization change unit 47 and a centralization change unit 48.

  The decentralization changing unit 47 includes the data allocation information 22 or the exclusive control information held by the data management unit 50 that implements the data allocation management function or the data exclusive control function among the data management units 50 operating in the centralized data management method. 24 is transmitted to another stubbed data management unit 50 that has been delegated to another without realizing the data arrangement management function or the data exclusive control function, and after the transmission is finished, the other stubbed data management unit 50, a data arrangement management function or a data exclusive control function is realized.

  This can be realized by performing the management information transmission plan and the operation of the method stop switching / resuming unit 38 in the method changing unit 30 in the first embodiment as follows.

  For example, if the method before and after the change is a distributed data management method, all storage nodes need to transmit data management information to each other. Further, the system stop switching / resuming unit 38 needs to match the stop and restart timings by the distributed synchronization algorithm.

  On the other hand, if the pre-change method is the centralized data management method and the post-change method is the distributed data management method, the storage node that is not stubbed by the centralized data management method becomes the source node and all storage The node corresponds to the destination node. Further, the system stop switching / resuming unit 38 needs to match the stop and restart timings by the distributed synchronization algorithm. Here, as an example of the distributed synchronization algorithm, the method described in Non-Patent Document 8 can be used.

  The centralization changing unit 48 transmits the data arrangement information 22 and the exclusive control information 24 in at least one storage node in which the data management unit 50 operates in the distributed data management method to the data management unit 50 of the other storage node, and the transmission ends Then, the data management unit is made a stub data management unit by delegating the data arrangement management function to another storage node.

  This can be realized by performing the management information transmission plan and the operation of the method stop switching / resuming unit 38 in the method changing unit 30 in the first embodiment as follows.

  When the method before the change is the distributed data management method and the method after the change is the centralized data management method, only the storage node that is stubbed by the centralization transmits the data management information. Further, the method stop switching / resuming unit 38 can stub each storage node without synchronizing.

  The storage system according to the present embodiment can provide an exclusive control method according to a usage trend that changes with time for each data set, and thus the system performance of the storage system can be improved. In addition, according to the storage system according to the present embodiment, it is possible to provide a data arrangement management method according to a usage trend that changes with time for each data set, so that it is possible to improve resource utilization efficiency in the storage system. it can.

  The reason is that the data set identification unit 51 identifies the data set to be accessed, and different data management methods of distribution and concentration can be applied to each data set. For data sets with high read access frequency and more efficient participation of many storage nodes, distributed data management with many high-speed storage nodes participating, and centralized management for other data sets Therefore, it is possible to improve the access performance for frequently used data sets.

  Further, the data change method can be dynamically changed for each data set by the method change unit 30. At this time, when the use of the first data set is completed and the second data set is next accessed many times, the first data set operating in the distributed data management method is changed to the centralized management method. By setting the second data set as the distributed data management method, it is possible to provide a data arrangement management method corresponding to the usage tendency for temporal conversion for each data set.

  When the method is changed, the access speed for the second data set can be further improved by relocating the data management function of the second data set to a higher-performance storage node.

  Further, in the storage node that has received access from the user computer 90 that has used the first data set, the data management function is stubbed and the access is passed to other storage nodes. Therefore, even when the data management method is changed, from the user computer 90 side, data can be used in the same manner as before except for performance.

(Third embodiment)
A storage system according to a third embodiment of the present invention will be described with reference to the drawings. In the present embodiment, description will be given of operations in consideration of rearrangement of stored data and cooperative operation of the present invention.

  In this embodiment, the storage node 70 includes a high-speed and low-capacity storage node having a high-speed and low-capacity data storage device 60 and a low-speed and large-capacity storage node having a low-speed and large-capacity data storage device 60. The storage node 70 may be classified according to the capacity of the data storage device 60 used and the type of device (for example, a magnetic disk or a non-volatile memory) that implements the data storage device 60. However, the classification method is not limited to these.

  FIG. 8 is a block diagram showing the configuration of the storage system according to this embodiment. Referring to FIG. 8, the storage system of the present embodiment further includes a data migration execution unit 77 and a management method change unit 78 in addition to the storage system of the first embodiment (see FIG. 1).

  The data movement execution unit 77 stores data sets that are more frequently used in a faster storage based on data usage trends (for example, the number of recently accessed accesses and the total size of accessed data) during system operation. Move to a node and move a less frequently used data set to a slower storage node. Thereby, resources in the storage system can be used effectively.

  When the data movement execution unit 77 moves the data set, the management method change unit 78 uses the method change unit 30 to change the data management method of the movement target data set. The management method changing unit 78 uses the first data set distributed and stored in at least one high-speed and low-capacity storage as a distributed data management method, and the data movement execution unit moves the first data set to the low-speed and large-capacity storage. First, the data management unit corresponding to the first data set is set to a centralized data management method (data degeneration function). However, the data management function that is not stubbed in the centralized data management method is arranged on a high-speed and low-capacity storage.

  In addition, the management method changing unit 78 uses the second data set distributed and stored in at least one low-speed large-capacity storage node as a centralized data management method, and some of the data management functions that are not stubbed are high-speed and low-capacity storage. When the second data set is moved to the high-speed and low-capacity storage by the data movement execution unit 78 placed in the node, the data management unit corresponding to the second data set is set as a distributed data management method (data expansion function) .

  By using the data reduction function and the data expansion function, the data management function is always located on the high-speed and low-capacity storage, and the response performance to the user computer can be maintained. Furthermore, by changing the method of the data management function in accordance with the data rearrangement process that reflects the data usage mode, it is possible to provide a data management function that matches the usage trend of the data set.

(Fourth embodiment)
A storage system according to the fourth embodiment of the present invention will be described with reference to the drawings. In this embodiment, power consumption in the storage system is reduced. FIG. 9 is a block diagram showing the configuration of the storage node in the storage system according to this embodiment. Referring to FIG. 9, a part or all of the storage node 70 in the storage system according to the present embodiment is more specific to the configuration of the storage node 70 of the first embodiment (see FIG. 2). An execution unit 75 and a power control unit 76 are provided.

  The specific data management method execution unit 75 can execute at least one data management method (hereinafter referred to as “low power consumption data management method”) with low power consumption, and the storage node 70 can reduce power consumption in its own node. When only the power data management method is operating, the CPU 71, the memory 72, the I / O controller 73, or a combination thereof is set to the low power consumption mode or stopped.

  The power control unit 76 monitors the data management method that operates in its own node, and when the operation of the data management method other than the low power consumption data management method is completed by the method change unit 30, the CPU 71, the memory 72, the I / O The O controller 73 or a combination thereof is set to the low power consumption mode or stopped. Whether the CPU 71, the memory 72, the I / O controller 73, or a combination thereof is operating in the low power consumption mode before the method changing unit 30 starts the operation of the data management method other than the low power consumption data management method. Or, when stopped, return to the normal operating state.

  As the specific data management method execution unit 75, for example, a stub node of a centralized data management method can be mounted on a network card. Processing without using the CPU 71 or the memory 72 of the storage node 70 by placing hardware having a function of transferring the received packet as it is to a device having a specified data management function on the network card. Can do.

  Further, the specific data management method execution unit 75 may be implemented by, for example, an ASIC that implements a specific data management method or a variable logic such as an FPGA. However, the implementation of the specific data management method execution unit 75 is not limited to these.

  By providing the specific data management method execution unit 75 and the power control unit 76, the power consumption can be further reduced along with the conversion of the data management method. This is because the specific data management method execution unit 75 can reduce the amount of power used when only the data management method that can be operated with low power consumption operates. In particular, when the amount of access to a large number of data sets stored in the storage system is small, the data management method for the large number of data sets can be made a centralized data management method. According to the centralized management data method, many storage nodes are stubbed. By introducing a function that can provide only the functions of the stubbed node with low power consumption, the storage system can The power consumption of the storage node can be suppressed without providing a function of rejecting access or causing the user computer to search for a data management function.

  The storage system according to the present invention can be applied to storage systems such as distributed storage, parallel file systems, RAID arrays, distributed databases, data grids, and cloud data storage platforms. In particular, according to the storage system of the present invention, it is possible to improve the performance of a storage system that executes user processing with a high-speed and low-capacity data grid and holds user data with low-speed and large-capacity storage.

  Although the above description has been made based on the embodiment, the present invention is not limited to the above embodiment. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

10, 10a to 10c Data management device 20 Data management method execution unit 22 Data arrangement information 24 Exclusive control information 30 Method change unit 31 Still point information generation unit 32 Management difference information generation holding unit 33 Still point information transmission unit 34 Still point information reception Unit 35 management difference information transmission unit 36 management difference information reception unit 37 management difference information application unit 38 method stop switching resumption unit 40 method selection unit 47 decentralization change unit 48 centralization change unit 50, 50a to 50c data management unit 51 data set Identification unit 60, 60a-60c Data storage device 70, 70a-70c Storage node 71 CPU
72 Memory 73 I / O Controller 74 Communication Controller 75 Specific Data Management Method Execution Unit 76 Power Control Unit 77 Data Movement Execution Unit 78 Management Method Change Unit 80 Network 90 User Computer 91 Data Utilization Process

Claims (18)

A data storage device storing a plurality of data sets;
A data management device having a data management unit corresponding to each of the plurality of data sets;
The data management unit includes a data placement management function for searching for data included in a corresponding data set, and a data exclusive control function for temporarily stopping access to data included in the corresponding data set. A data management method execution unit realized based on one of the methods;
The data management method execution unit further includes a method change unit that changes the data management method from the first data management method to the second data management method among the plurality of data management methods. Storage node. Each of the plurality of data sets includes data identified by a data identifier;
The data management device further comprises a data set identification unit that returns a data set identifier that identifies a data set to which the data belongs among the plurality of data sets when receiving access to data from a data utilization process,
The data management unit searches the data storage device in which the data is stored by referring to the data arrangement information indicating the correspondence between the data and the data storage device in which the data is stored and the data identifier, The storage node according to claim 1, wherein the storage node implements the data arrangement management function.   The data management unit realizes the data exclusive control function by referring to exclusive control information for stored data and temporarily stopping access to the data by a data use process. Item 3. The storage node according to Item 1 or 2. The data management unit further includes a method selection unit that generates a data management method identifier representing a data management method from data management method information,
The data management method execution unit realizes the data placement management function and the data exclusive control function based on a data management method specified by the data management method identifier of the plurality of data management methods. The storage node according to any one of claims 1 to 3.   5. A storage system comprising a plurality of storage nodes according to claim 1, connected to each other via a network. The plurality of data management methods include at least distributed data that realizes the data placement management function and the data exclusive control function by exchanging information between the distributed data management functions of a plurality of storage nodes via a network. Management method,
A centralized data management method in which at least one storage node is stubbed by delegating a request for these functions to another storage node without realizing the data arrangement management function or the data exclusive control function The storage system according to claim 5, comprising: The method change unit transmits data arrangement information and exclusive control information in at least one storage node that operates in the distributed data management method to the data management unit of another storage node, and after the transmission ends, A centralized change unit that stubs the data management unit by delegating the data placement management function to another storage node;
Among the data management units that operate in the centralized data management method, a part of the data allocation information or exclusive control information held by the data management unit that implements the data allocation management function or the data exclusive control function is converted into a stub. Further comprising: a decentralized changing unit that transmits the data to the other data management unit and, after the transmission is completed, causes the other data management unit to realize the data arrangement management function or the data exclusive control function. Item 6. The storage system according to Item 5. The method changing unit generates a stationary point information generating unit that generates the data location information of the transmission source storage node at a specific time and the value of the exclusive control information as stationary point information;
A quiesce point information transmission unit for transmitting quiesce point information to other storage nodes;
A quiesce point information receiving unit that receives and holds quiesce point information transmitted from other storage nodes;
A management difference information generating and holding unit that compares the value at the current time with access from the user computer and the static point information, and generates and holds the difference between the static point information and the value at the current time as management difference information;
A management difference information transmission unit for transmitting management difference information to another storage node;
A management difference information receiving unit for receiving management difference information transmitted from another storage node;
Based on the still point information at the first time and the management difference information from the first time to the second time, the data arrangement information and the exclusive control information at the second time are restored as new still point information. A management difference information application unit;
The quiesce point information generating unit generates quiesce point information, transmits the value of the quiesce point information to the transmission destination storage node, and changes to the data arrangement information being transmitted and the exclusive control information as the first management difference information, the transmission source storage When the management difference information exists in the management difference information generation holding unit in the transmission source storage after the transmission of the still point information value is held in the management difference information generation holding unit in the node, the transmission difference storage node stores the management difference information. Management difference information receiving unit, and the management difference information applying unit in the storage node of the transmission destination reproduces the management difference information at the time of transmission in the transmission destination storage based on the received management difference information and the received still point information. In operation, the size of the management difference information is equal to or smaller than the predetermined size, and the number of transmissions of the management difference information is equal to or larger than the predetermined number. 8. The method further comprising: a method stop switching / resuming unit that changes the data management method in the data management method execution unit and transmits the remaining management difference information The storage system according to any one of the above.   9. The storage system according to claim 8, wherein the centralized changing unit stubs the data arrangement management function in the method stop switching / resuming unit. The data set identification unit further includes an access waiting unit uniquely associated with each of the plurality of data sets;
The data use process accesses an access waiting unit corresponding to a data set to which data to use belongs,
The storage system according to any one of claims 5 to 9, wherein the data set identification unit identifies a data set based on an access waiting unit that has received an access.   The storage system according to any one of claims 5 to 10, wherein the data set identification unit specifies a data set by receiving data set identification information together with access from the data use process.   The data set identification unit holds data / data set correspondence information that can identify a data set from a data identifier, and includes a data identifier included in an access from a data utilization process and the data / data set correspondence information. The storage system according to any one of claims 5 to 11, wherein a data set is specified based on the data set.   The data set identification unit holds client / data set correspondence information that can identify a data set from a data use process or user identification information that specifies a user computer in which the data use process operates. The storage system according to any one of claims 5 to 12, wherein the data set is specified based on user identification information obtained by receiving access and the client data set correspondence information. The plurality of storage nodes include a high-speed and low-capacity storage node having a high-speed and low-capacity data storage device, and a high-speed and low-capacity storage node having a low-speed and large-capacity data storage device,
When the data set distributed and stored in the high-speed and low-capacity storage is the distributed data management method, and the data set is moved to the low-speed and large-capacity storage, the data management unit corresponding to the data set is The storage system according to any one of claims 5 to 13, further comprising: a management method changing unit that uses the data management method.   The management method changing unit uses the centralized data management method for data sets distributed and stored in the low-speed and large-capacity storage, and corresponds to the data set when the data set is moved to the high-speed and low-capacity storage. 15. The storage system according to claim 14, wherein the data management unit to be used is the distributed data management method. The storage node includes a specific data management method execution unit that executes a predetermined data management method of the plurality of data management methods with low power consumption;
A general data management method execution unit for executing a data management method other than the predetermined management method;
When a data management method other than the predetermined management method is not operating in its own node, it further comprises a power control unit that shifts the general data management method execution unit to a low power consumption state. The storage system according to any one of claims 5 to 15. A data management unit corresponding to each of the plurality of data sets stored in the storage data storage device,
A data placement management step of searching for data contained in a corresponding data set based on one of a plurality of data management methods;
A data exclusive control step of temporarily stopping access to data included in a corresponding data set based on any of the plurality of data management methods;
And a step of changing a data management method in the data arrangement management step and the data exclusive control step from a first data management method of the plurality of data management methods to a second data management method. Data storage method. For the CPU corresponding to each of the plurality of data sets stored in the data storage device of the storage,
A data placement management process for searching for data contained in a corresponding data set based on one of a plurality of data management methods;
A data exclusive control process for temporarily stopping access to data included in a corresponding data set based on one of the plurality of data management methods;
A process of changing a data management method in the data arrangement management process and the data exclusive control process from a first data management method of the plurality of data management methods to a second data management method is executed. Program to do.

Images