JP6076882B2 - Information processing system, management device, and key assignment program - Google Patents

Description

  The present invention relates to an information processing system, a management device, a server device, and a key assignment program.

  Conventionally, KVS (Key Value Store) is known as one of data storage and management techniques in computer systems. In KVS, a unique indicator (key) corresponding to arbitrary data (value) to be stored is set, and the value and key are stored in pairs. There is also known a distributed KVS system having a function of storing data in a distributed manner in a plurality of servers and storage devices.

  In such distributed KVS, in general, a server that stores data of a key is determined from the hash value of the key according to a consistent hash method or the like. Due to the randomness of the hash function, each server is assigned an approximately equal number of keys.

JP2013-15269A

David Karger, Eeic Lehman, Tom Leighton, Matthew Levine, Daniel Lewin, and Rina Panigrahy, "Consistent Hashing and Random Trees: Distributed Caching Protocols for Relieving Hot Spots on the World Wide Web" in Proceedings of the 29th annual ACM Symposium on Theory of Computing (STOC'97), pp.654-663, May 1997.

  However, in the conventional distributed KVS system technology, when access concentrates on a specific key, there is a problem that access to the server to which the key is assigned is concentrated and the server is congested. In addition, there is a problem that resources of the entire system cannot be effectively used because access is concentrated on some servers.

  Therefore, an object of the present invention is to reduce server congestion and effectively use resources of the entire system.

  In order to solve the above-described problems and achieve the object, an information processing system disclosed includes a plurality of storage devices that store data and a key for identifying each data in association with each other, and access to the plurality of storage devices In an information processing system having a server device that performs the above and a management device that assigns the key to data stored in a storage device, when the server device accepts an access request including the key, the key An access unit that accesses a storage device that stores data corresponding to the above, and a counting unit that counts the number of accesses, which is the number of access requests received, for each key, and the management device includes: An acquisition unit that acquires the access count for each key from the server device, and the access count acquired by the acquisition unit exceeds a predetermined threshold A determination unit that determines the number of copies of the key corresponding to the access request, a creation unit that creates a copy of the number of keys determined by the determination unit, and a copy of the key created by the creation unit, And an allocating unit that allocates data stored in any one of the plurality of storage devices to data corresponding to a key whose number of accesses exceeds the threshold value.

  Further, the disclosed management device includes an acquisition unit that acquires the number of accesses for each key from a server device that accesses a plurality of storage devices that store data and a key that identifies each data in association with each other; When the number of accesses acquired by the acquisition unit exceeds a predetermined threshold, a determination unit that determines the number of copies of the key corresponding to the access request, and the number of copies of the key determined by the determination unit And a copy of the key created by the creating unit is data stored in any one of the plurality of storage devices, and the access count exceeds the threshold value An allocating unit that allocates data corresponding to the key.

  In addition, when the disclosed server device receives an access request including the key for a plurality of storage devices that store the data and a key for identifying each data in association with each other, the server device of the access request A counting unit that counts the number of accesses that is a number, and when an access request including the key is received, the key is a duplicate key of a key whose number of accesses counted by the counting unit exceeds a predetermined threshold An access unit that selects one key from among the duplicated keys and uses the key to access the storage device. It is characterized by providing.

  Further, the disclosed key assignment program includes an acquisition step of acquiring the number of accesses for each key from a server device that accesses a plurality of storage devices that store data and a key that identifies each data in association with each other, When the number of accesses acquired by the acquisition step exceeds a predetermined threshold, a determination step for determining the number of copies of the key corresponding to the access request, and a number of keys determined by the determination step A creation step of creating a copy, and a copy of the key created by the creation step is data stored in any one of the plurality of storage devices, and the access count exceeds the threshold value An assigning step for assigning data corresponding to the key is executed by a computer.

  The information processing system, management device, server device, and key assignment program disclosed in the present application can reduce server congestion and effectively utilize resources of the entire system.

FIG. 1 is a diagram illustrating an example of a configuration of an information processing system according to the first embodiment. FIG. 2 is a diagram illustrating an overview of the information processing system according to the first embodiment. FIG. 3 is a block diagram showing the configuration of the request processing server according to the first embodiment. FIG. 4 is a diagram illustrating an example of information stored in the request frequency storage unit according to the first embodiment. FIG. 5 is a diagram illustrating an example of information stored in the replication management information table storage unit according to the first embodiment. FIG. 6 is a block diagram showing the configuration of the replication management server according to the first embodiment. FIG. 7 is a diagram illustrating an example of information stored in the request frequency information table storage unit according to the first embodiment. FIG. 8 is a diagram for explaining key assignment processing in the replication management server according to the first embodiment. FIG. 9 is a sequence diagram for explaining the flow of the service request response process in the information processing system according to the first embodiment. FIG. 10 is a sequence diagram for explaining the flow of the copy number changing process in the information processing system according to the first embodiment. FIG. 11 is a flowchart for explaining the flow of the replication number change determination process in the replication management server according to the first embodiment. FIG. 12 is a diagram for explaining an example in which accesses are concentrated on a specific key and loads are concentrated on a distributed KVS server to which the key is assigned. FIG. 13 is a block diagram illustrating a configuration of a replication management server according to the second embodiment. FIG. 14 is a diagram illustrating an example of information stored in the replication management information table storage unit according to the second embodiment. FIG. 15 is a diagram illustrating an example of information stored in the request frequency information table storage unit according to the second embodiment. FIG. 16 is a sequence diagram for explaining the flow of the copy number changing process in the information processing system according to the second embodiment. FIG. 17 is a diagram illustrating a computer that executes a key assignment program.

  Exemplary embodiments of an information processing system, a management device, a server device, and a key assignment program according to the present invention will be described below in detail with reference to the drawings. In addition, this invention is not limited by this embodiment.

[First embodiment]
In the following embodiments, the configuration of the information processing system according to the first embodiment, the configuration of the request processing server, the configuration of the replication management server, and the processing flow in the information processing system will be described in order, and finally the first embodiment The effect of will be described.

[System configuration]
First, an example of the configuration of the information processing system 1 according to the first embodiment will be described. FIG. 1 is a diagram illustrating an example of a configuration of an information processing system according to the first embodiment. As shown in FIG. 1, the information processing system 1 includes a plurality of request processing servers 10A to 10E, a replication management server 20, a plurality of clients 30A to 30C, a load distribution device 40, a plurality of distributed KVS servers 50A to 50E, a plurality of IP networks 60A-60C are provided.

  Note that the request processing servers 10A to 10E, the clients 30A to 30C, the distributed KVS servers 50A to 50E, and the IP networks 60A to 60C are described without particular distinction, “request processing server 10”, “client 30”. , “Distributed KVS server 50” and “IP network 60”. Moreover, the installation number of each apparatus illustrated in FIG. 1 is an example to the last, and is not limited to this.

  As shown in FIG. 1, the clients 30A to 30C and the load balancer 40 are connected via an IP network 60A. Further, the load balancer 40, the plurality of request processing servers 10A to 10E, and the replication management server 20 are connected via an IP network 60B. The plurality of request processing servers 10A to 10E and the replication management server 20 are connected to the plurality of distributed KVS servers 50A to 50E via the IP network 60C.

  The clients 30 </ b> A to 30 </ b> C are general-purpose terminals that transmit service requests to the load balancer 40 and provide various services to the user using service responses returned from the load balancer 40.

  The load distribution device 40 is a general-purpose load distribution device that distributes service requests transmitted from the clients 30A to 30C to the request processing servers 10A to 10E in the order of arrival.

  The distributed KVS servers 50A to 50E are general-purpose server devices on which server software of a general-purpose distributed KVS system operates. In the distributed KVS servers 50A to 50E, data (value) and a key for identifying each data are stored in association with each other.

  The request processing servers 10A to 10E access the distributed KVS servers 50A to 50E. Here, the outline of the information processing system according to the first embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating an overview of the information processing system according to the first embodiment. Note that FIG. 2 illustrates an example in which three request processing servers 10A to 10C are installed and four distributed KVS servers 50A to 50D are installed. In FIG. 2, the client 30 and the IP network 60 are not shown.

  As shown in FIG. 2, the request processing server 10 is a general-purpose server device on which each software of the request processing application 110, the access conversion wrapper 120, and the distributed KVS client library 130 operates, and the software communicates by inter-process communication. .

  The request processing application 110 is general-purpose application software that creates a service response to a service request transmitted from the client 30, and is necessary for creating a service response by accessing the distributed KVS server 50 via the access conversion wrapper 120. The correct data.

  The access conversion wrapper 120 is software having an access unit 12a, which will be described later. The access conversion wrapper 120 provides an interface equivalent to the distributed KVS client library 130 to the request processing application 110, and the reference request and update transmitted from the request processing application 110. With respect to the request and the deletion request, it is determined whether the target key is duplicated according to the duplication management information notified from the duplication management server 20. If the key is duplicated, the request is converted and transmitted to the distributed KVS client library 130. .

  Further, the access conversion wrapper 120 includes a counting unit 12b described later, and notifies the replication management server 20 of the request frequency for each target key with respect to the reference request transmitted from the request processing application 110.

  The distributed KVS client library 130 is a general-purpose distributed KVS system client software, and for each request transmitted from the access conversion wrapper 120, determines a distributed KVS server 50 as a distribution destination, and transmits the request.

  The replication management server 20 determines the number of copies of each key in accordance with the request frequency information transmitted from the request processing servers 10A to 10C, and creates and creates as many replication keys as the determined number of copies for the distributed KVS server 50. The duplicate key is assigned to the distributed KVS servers 50A to 50D. Also, the distributed KVS client library 210 exists on the replication management server 20, and a reference request and an add request, which will be described later, and a response to the request are received via the distributed KVS client library 210.

[Configuration of request processing server]
Next, the configuration of the request processing server 10 shown in FIG. 1 will be described with reference to FIG. FIG. 3 is a block diagram showing the configuration of the request processing server according to the first embodiment. As illustrated in FIG. 3, the request processing server 10 includes a communication processing unit 11, a control unit 12, and a storage unit 13.

  The communication processing unit 11 controls communication related to various types of information exchanged with the connected replication management server 20, load distribution device 40, and distributed KVS server 50. For example, the communication processing unit 11 transmits a request frequency for each reference target key to the replication management server 20 and receives a replication management information table from the replication management server 20.

  The communication processing unit 11 receives a service request from the load balancer 40 and transmits a service response that is a response to the service request to the load balancer 40. Further, the communication processing unit 11 transmits a reference request, an update request, and a deletion request to the distributed KVS server 50, and receives a reference response, an update response, and a deletion response corresponding to each request from the distributed KVS server 50. To do.

  As shown in FIG. 3, the storage unit 13 includes a request frequency storage unit 13a and a replication management information table storage unit 13b. The storage unit 13 is, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk.

  The request frequency storage unit 13a stores the request frequency of each reference target key. Specifically, the request frequency storage unit 13a stores a request frequency that is the number of access requests counted by a counting unit 12b described later for each reference target key.

  Here, an example of information stored in the request frequency storage unit 13a will be described with reference to FIG. FIG. 4 is a diagram illustrating an example of information stored in the request frequency storage unit according to the first embodiment. As illustrated in FIG. 4, the request frequency storage unit 13 a includes a “reference target key” that is a character string for uniquely identifying a reference target key, and the number of times each reference target key is an access request target. “Request frequency” is stored in association with each other. The value of the request frequency is a request frequency within a predetermined time (for example, 5 minutes), and the value is reset every 5 minutes.

  For example, the request frequency storage unit 13a stores “40” as the request frequency corresponding to the reference target key “aaa” and the request frequency corresponding to the reference target key “bbb”. “75” is stored, and “150” is stored as the request frequency corresponding to the reference target key “ccc”.

  The replication management information table storage unit 13b stores a replication management information table created by the replication management server 20. Specifically, the duplicate management information table storage unit 13b stores the number of duplicated keys for each duplicated reference target key as a duplicate management information table.

  Here, an example of information stored in the replication management information table storage unit 13b will be described with reference to FIG. FIG. 5 is a diagram illustrating an example of information stored in the replication management information table storage unit according to the first embodiment. As illustrated in FIG. 5, the replication management information table storage unit 13 b indicates a “key” that is a character string for identifying a reference target key replicated by the replication management server 20 and the number of replicated reference target keys. The “number of copies” is stored in association with each other.

  For example, the replication management information table storage unit 13b stores “3” as the number of replicas corresponding to the key “ddd” and “6” as the number of replicas corresponding to “ggg”. And “14” is stored as the number of copies corresponding to “hh”.

  Returning to FIG. 3, the control unit 12 includes an access unit 12a and a counting unit 12b. Here, the control unit 12 is an electronic circuit such as a CPU (Central Processing Unit) or MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

  When the access unit 12a receives an access request including a reference target key, the access unit 12a accesses the distributed KVS server 50 that stores data corresponding to the reference target key. Specifically, when the access unit 12a receives an access request including a reference target key, the access unit 12a determines whether the key is a duplicated key. One key is selected from the keys, and the distributed KVS server 50 is accessed using the key.

  For example, the access unit 12a performs the reference, update, and delete requests transmitted from the request processing application 110 to the distributed KVS server 50 based on the replication management information created by the replication management server 20. It is determined whether or not the key is duplicated, and if it is a request for the duplicated key, the reference request is converted and transmitted to the distributed KVS server 50.

  For example, when the access unit 12a is a request for a duplicated key, the reference request is converted into a reference request for a key randomly selected from the duplicate keys, and all of the update request and the delete request are converted. It is converted into an update request and delete request for the duplicate key.

  The counting unit 12b counts the number of accesses that is the number of received access requests for each reference target key. Specifically, the counting unit 12b counts the request frequency for each target key included in the reference request transmitted from the request processing application 110, and stores it in the request frequency storage unit 13a. Further, the counting unit 12b periodically transmits information stored in the request frequency storage unit 13a to the replication management server 20 as a request frequency information notification. For example, the counting unit 12b transmits a request frequency information notification to the replication management server 20 every 5 minutes. Further, when transmitting the request frequency information notification, the counting unit 12b resets the value of the request frequency stored in the request frequency storage unit 13a.

[Configuration of replication management server]
Next, the configuration of the replication management server 20 shown in FIG. 1 will be described with reference to FIG. FIG. 6 is a block diagram showing the configuration of the replication management server according to the first embodiment. As illustrated in FIG. 6, the replication management server 20 includes a communication processing unit 21, a control unit 22, and a storage unit 23.

  The communication processing unit 21 controls communication related to various information exchanged between the request processing server 10 and the distributed KVS server 50 to be connected. For example, the communication processing unit 21 receives the request frequency for each reference target key from the request processing server 10 and transmits a replication management information table to the replication management server 20. Further, the communication processing unit 21 transmits the created duplicate key to the KVS server 50.

  As shown in FIG. 6, the storage unit 23 includes a replication management information storage unit 23a and a request frequency information table storage unit 23b. The storage unit 23 is, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk.

  The replication management information table storage unit 23a stores a replication management information table created by the creation unit 22c described later. Specifically, the duplicate management information table storage unit 23a stores the number of duplicated keys for each duplicated reference target key as a duplicate management information table. Note that the replication management information table storage unit 23a stores the same information as the information stored in the replication management information table storage unit 13b described above. As illustrated in FIG. A “key” that is a character string to be identified and a “duplicate number” indicating the number of duplicated reference target keys are stored in association with each other.

  The request frequency information table storage unit 23b stores the number of copies of each reference target key together with the request frequency for each reference target key. Specifically, the request frequency information table storage unit 23b stores the number of copies of the reference target key determined by the determination unit 22b described later together with the request frequency for each reference target key.

  Here, an example of information stored in the request frequency information table storage unit 23b will be described with reference to FIG. FIG. 7 is a diagram illustrating an example of information stored in the request frequency information table storage unit according to the first embodiment. As illustrated in FIG. 7, a “reference target key” that is a character string for identifying a reference target key, a “request frequency” that is the number of times each reference target key is an access request target, and a target reference The “replication target M”, which is the number of copies of the target key, and the “replication number N”, which is the number of reference target keys that are actually copied, determined by the determination unit 22b described later, are stored in association with each other.

  To describe with a specific example, the request frequency information table storage unit 23b, for example, “40” as the request frequency corresponding to the reference target key “aaa”, “0” as the replication target M, and “ “0” is stored. Further, for example, the request frequency information table storage unit 23b stores “610” as the request frequency corresponding to the reference target key “ddd”, “2” as the replication target M, and “3” as the number of replications N.

  Returning to FIG. 6, the control unit 22 includes an acquisition unit 22a, a determination unit 22b, a creation unit 22c, and an allocation unit 22d. Here, the control unit 22 is an electronic circuit such as a CPU (Central Processing Unit) or MPU (Micro Processing Unit), or an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

  The acquisition unit 22a acquires the access count for each reference target key from the request processing server 10. Specifically, the acquisition unit 22a acquires a request frequency for each reference target key for a reference request from the request processing application 110 of the request processing server 10 to the distributed KVS server 50. Then, the acquisition unit 22a stores the acquired request frequency in the request frequency information table storage unit 23b.

  The determining unit 22b determines the number of copies of the reference target key corresponding to the access request when the number of accesses acquired by the acquiring unit 22a exceeds a predetermined threshold. Specifically, the number of copies of each reference request key is determined from the request frequency information table storage unit 23b according to the request frequency for each reference request key, and stored in the request frequency information table storage unit 23b.

  In addition, when the number of accesses acquired by the acquisition unit 22a exceeds the threshold, the determination unit 22b determines the number according to a value obtained by dividing the number of accesses by the threshold. Here, the process of determining the number of copies of each reference request key will be described in detail. For example, the determination unit 22b adds the request frequencies acquired by the acquisition unit 22a for each target key, creates a request frequency information table as illustrated in FIG. 7 and stores the request frequency information table in the request frequency information table storage unit 23b. To do.

  Here, the determination unit 22b divides the request frequency of each key by a predetermined threshold value, and calculates an integer obtained by rounding off the decimal point as the replication target M in the request frequency information table. For example, when the threshold value is “300”, the replication target M of the key “ddd” whose request frequency is 610 is “2”.

  The replication target M represents the number of servers to which the key should be replicated in order to reduce the load bias of the distributed KVS server 50. For example, the replication target M of the key “ddd” whose replication target M is “2” In this case, it means that it is necessary to create a duplicate key of the key “ddd” for two distributed KVS servers different from the distributed KVS server 50 to which the key “ddd” is originally assigned.

  However, the distributed KVS server 50 to which the key is assigned is determined by the distributed KVS client library 130 in a random manner by, for example, the consistent hash method. Cannot be specified. For this reason, even if replication keys for the replication target M are created, they are not all assigned to different distributed KVS servers 50.

  Therefore, when a replication key is randomly assigned to each distributed KVS server 50, an expected value of the number of replications required for the number of distributed KVS servers 50 to which a replication key is assigned to be equal to or greater than the replication target M is the replication number N. Define. The determination unit 22b calculates the number of copies N by the following equation (1), and rounds the number after the decimal point to make it an integer. In the following (1), the number of distributed KVS servers 50 is S.

  It should be noted that the state in which one copy key exists in each of all servers other than the distributed KVS server to which a certain key is originally assigned is the state in which the load can be distributed most. Therefore, the maximum value of the replication target M is “S−1”. " The above-described request frequency information table illustrated in FIG. 7 is an example when the number of servers S of the distributed KVS is “10”.

  Thereafter, the determination unit 22b extracts one having a copy count of 1 or more from the created request frequency information table, and creates a copy management information table as illustrated in FIG. For example, when there is a replication management information table created last time, the determination unit 22b compares it with the replication management information table created this time, and records with an increased number of replications (including newly added records). , Records with a reduced number of replicas (including deleted records) are extracted.

  Further, if the replication management information table created at the previous time does not exist, the determination unit 22b makes all the records of the replication management information table created this time a record in which the number of replications has increased. Note that the replication management information table created last time is deleted because it is unnecessary after the comparison of the replication management table.

  The creation unit 22c creates as many copies of the reference target keys as the number determined by the determination unit 22b. Specifically, the creation unit 22c creates, for each reference target key, as many replication keys as the number of copies stored in the request frequency information table storage unit 23b, and creates replication management information to create a replication management information storage unit. 23a.

  For example, the creation unit 22c adds a predetermined delimiter (for example, “_” (underscore)) to the key character string (for example, “abc”) and 1 to “the copy of the key acquired from the copy management information. Character strings (for example, “abc_1”, “abc_2”, “abc_3”) to which integers up to “number” (for example, “1” to “3”) are assigned are created as new duplication keys.

  The allocating unit 22d is a reference stored with the copy of the reference target key created by the creating unit 22c stored in any one of the plurality of distributed KVS servers, and the access count exceeds the threshold. Assign to the data corresponding to the target key. The specific contents of the process assigned to the data corresponding to the reference target key will be described later as a duplicate key adding process in FIG.

  Here, the outline of the key assignment processing in the replication management server 20 will be described using a specific example with reference to FIG. FIG. 8 is a diagram for explaining key assignment processing in the replication management server according to the first embodiment. As illustrated in FIG. 8, the replication management server 20 replicates a key (for example, keyY) that is accessed frequently according to the reference request frequency of each key, and replicates keys (for example, keyY_1 to 3). Are assigned to the distributed KVS servers 50A, 50C, and 50D.

  For example, the request processing server 10B converts a reference request for the key Y, which is a high-frequency access key, into a reference request for the duplicate key (key Y_2), and accesses the distributed KVS server 50C.

  As described above, in the information processing system 1 according to the first embodiment, the replication management server 20 creates a copy of each key according to the request frequency of each key in the reference request from the application, and the reference request from the application. Are distributed over these replicated keys. For this reason, while using the existing distributed KVS system as it is and without requiring a special function addition to the request processing application, the load unevenness resulting from the uneven access frequency of the keys in the distributed KVS server 50 Can be reduced.

[Processing by information processing system]
Next, processing by the information processing system 1 according to the first embodiment will be described with reference to FIGS. 9 and 10. FIG. 9 is a sequence diagram for explaining the flow of the service request response process in the information processing system according to the first embodiment. FIG. 10 is a sequence diagram for explaining the flow of the copy number changing process in the information processing system according to the first embodiment.

  First, the flow of service request response processing in the information processing system 1 will be described with reference to FIG. In FIG. 9, processing until the request processing server 10 returns a service response to the service request transmitted from the client 30 will be described.

  As shown in FIG. 9, the client 30 transmits a service request to the load balancer 40 (step S101). Then, the load distribution apparatus 40 that has received the service request determines the request processing server 10 to which the service request is transferred by a method such as round robin, and transfers the service request (step S102). Here, it is assumed that the service request includes a user name, a service specification value, a service option specification value, and the like.

  Subsequently, the request processing application 110 on the request processing server 10 that has received the service request acquires, from the distributed KVS server 50, information necessary for creating a service response to the service request. Is transmitted to the access conversion wrapper 120 (step S103). Although not described in FIG. 9, when it is necessary to acquire other data for creating a service response, the request processing application 110 uses a key (for example, a user name) for acquiring the data. Or a service option specification value) is transmitted as a target key, and necessary data is acquired from the reference response.

  Subsequently, the access conversion wrapper 120 that has received the reference request transmits a reference request using the target key included in the reference request as a key to the distributed KVS client library 130 (step S104). Specifically, the access conversion wrapper 120 searches the replication management information table illustrated in FIG. 5 using the target key included in the reference request as a key. Then, if there is no corresponding record as a result of the search, the access conversion wrapper 120 determines that the key is not duplicated, and the access conversion wrapper 120 distributes the reference request transmitted from the request processing application as it is. Transmit to the client library 130.

  On the other hand, when the corresponding record exists, that is, when the key is duplicated, the access conversion wrapper 120 adds a predetermined delimiter (for example, “abc”) to the character string (for example, “abc”) of the key. _ ”(Underscore)) and a character string (for example,“ 3 ”) randomly selected from a range from 1 to“ the number of copies of the key acquired from the replication management information ”(for example,“ 3 ”) A reference request having “abc_3)) as a new target key is transmitted to the distributed KVS client library 130. Thus, the reference request for the duplicated key is a reference request for one of the duplicate keys (such as“ abc_3 ”). It is possible to reduce the load bias of the distributed KVS server 50.

  Then, the distributed KVS client library 130 that has received the reference request uses the target key included in the reference request to specify the distributed KVS server 50B that holds the data of the key by a method such as a consistent hash method, for example. Then, a reference request is transmitted to the distributed KVS server 50B (step S105).

  Then, the distributed KVS server 50B that has received the reference request obtains data corresponding to the target key included in the reference request from its own storage area, and sends a reference response storing the data to the distributed KVS client library on the request processing server. It transmits to 130 (step S106). If there is no data corresponding to the target key, an empty reference response is transmitted.

  Then, the distributed KVS client library 130 that has received the reference response transfers the reference response to the access conversion wrapper 120 (step S107). Subsequently, the access conversion wrapper 120 that has received the reference response transfers the reference response to the request processing application 110 (step S108).

  The request processing application 110 that has received the reference response creates a service response using the data included in the reference response, and transmits the service response to the client 30 (step S109). Here, for example, when it is necessary to update the data of the distributed KVS servers 50A to 50C in order to record the final request transmission time of each user who transmitted the service request, the request processing application 110 An update request having the name as a target key and the last request transmission time as update data is transmitted to the access conversion wrapper 120 (step S110).

  Then, the access conversion wrapper 120 that has received the update request transmits the update request to the distributed KVS client library 130 (step S111). Specifically, the access conversion wrapper 120 that has received the update request searches the replication management information table using the target key included in the update request in the same manner as when the reference request is received. If there is no corresponding record as a result of the search, that is, if the key is not duplicated, the access conversion wrapper 120 transmits the update request transmitted from the request processing application to the distributed KVS client library 130 as it is.

  On the other hand, if the corresponding record exists, that is, if the key is duplicated, the access conversion wrapper 120 adds a predetermined delimiter character (for example, “UserA”) to the character string (for example, “UserA”) included in the update request. For example, “_” (underscore)) and a character string (for example, “UserA_1”) to which 1 to “the number of copies of the key acquired from the replication management information” (for example, “1” to “5”) are added, (UserA_2, UserA_3, UserA_4, UserA_5)) and an update request using the copy source key “UserA” as update target keys are transmitted to the distributed KVS client.

  As a result, the copy source key and all copy keys are updated. Although not described in FIG. 9, even when a deletion request is received from the request processing, the deletion request for deleting the replication source key and all the replication keys is distributed as in the case of receiving the update request. Transmit to the KVS client library 130.

  Then, the distributed KVS client library 130 that has received these update requests uses the target key included in each update request to store the data of the key by a method such as a consistent hash method, as in the case of receiving the reference request. The distributed KVS servers 50A to 50C to be held are specified, and an update request is transmitted to the distributed KVS servers 50A to 50C (step S112).

  The distributed KVS servers 50 </ b> A to 50 </ b> C that have received the update request overwrite the data corresponding to the target key included in the update request with the data included in the update request, and update the update response to the distributed KVS client library 130 on the request processing server 10. (Step S113).

  Then, the distributed KVS client library 130 that has received the update response transfers the update response to the access conversion wrapper 120 (step S114). Subsequently, the access conversion wrapper 120 transmits an update response to the request processing application 110 after receiving update responses to all the update requests transmitted in step S111 (step S115).

  Next, the flow of the copy number changing process in the information processing system 1 will be described with reference to FIG. FIG. 10 illustrates a process in which the replication management server 20 determines the number of key replications according to the request frequency of each key transmitted from the access conversion wrapper 120, and adds and deletes replication keys. In the above processing, the replication management server 20 periodically performs processing in accordance with the interval at which the request processing server 10 transmits request frequency information.

  As shown in FIG. 10, the access conversion wrapper 120 on each request processing server 10 performs a request frequency counting process that counts the request frequency for each target key included in the reference request transmitted from the request processing application 110 (steps). S201).

  Then, the access conversion wrapper 120 periodically transmits the request frequency information as a request frequency information notification to the replication management server 20 (for example, every 5 minutes) (step S202). The request frequency information includes the request frequency for each target key. However, if the request frequencies for all target keys in the reference request received within the period are stored, the data volume can be significantly increased. Therefore, only the information on the target key having a request frequency exceeding a predetermined threshold (for example, the number of requests for 5 minutes is 100 or more) may be included in the request frequency information.

  Thereafter, the replication management server 20 that has received the request frequency information notification performs a replication number change determination process, which will be described in detail later with reference to FIG. 11 (step S203). Then, the replication management server 20 reflects the increase and decrease in the number of replications extracted in the replication number change determination process to the distributed KVS servers 50A to 50C as additions and deletions of the replication keys, respectively, and in the replication management information table. The update is notified to the access conversion wrapper 120. This process is performed in the order of a replication key addition process, a replication management information update notification, and a replication key deletion process. This is to prevent the access conversion wrapper 120 from accessing the duplicate key before addition or after deletion.

  In the replication key addition process, the replication management server 20 first makes a reference request for the replication source key (step S204), receives a reference response to the reference request (step S205), and receives data (value) corresponding to the key. ) To get. Although not shown in FIG. 11, the distributed KVS client library 210 also exists on the replication management server 20, and the transmission of reference requests and additional requests described below and the reception of responses to these requests are distributed. This is done via the KVS client library.

  Next, the replication management server 20 makes a key addition request for the increase in the number of replicas. For example, when the predetermined delimiter is “_” and the number of copies of the key ggg is increased from 3 to 6, a new duplicate key is added with “ggg_4”, “ggg_5”, and “ggg_6” as target keys. An addition request for processing is transmitted to the distributed KVS servers 50A to 50C (steps S206, S208, and S210), and an additional response is received from the distributed KVS servers 50A to 50C (steps S207, S209, and step S211). ).

  Subsequently, the replication management server 20 performs the replication key addition process for all keys whose number of replications has increased as described above, and then notifies the access conversion wrapper 120 of each request processing server of the replication management information update notification. (Step S212), and a copy management information update response is received as a response (Step S213). Note that the replication management information update notification does not need to include all the contents of the new replication management information table, only the update differences are stored, and the access conversion wrapper creates a new replication management information table to which the update differences are applied. May be.

  Then, the replication management server 20 sends a replication management information update notification to all the request processing servers 10, and then transmits a deletion request for requesting a replication key deletion process to the distributed KVS servers 50B and 50C (step S40). S214, step S216), and the response is received (step S215, step S217). Here, in the copy key deletion process, in contrast to the addition process, for example, when the number of copies of the key ggg is reduced from 8 to 6, the deletion requests with the target keys “ggg_8” and “ggg_7” are distributed. This is performed for the KVS servers 50B and 50C.

  Next, the above-described copy number change determination process will be described with reference to FIG. FIG. 11 is a flowchart for explaining the flow of the replication number change determination process in the replication management server according to the first embodiment. As shown in FIG. 11, the replication management server 20 creates a request frequency information table (step S301).

  Specifically, the replication management server 20 adds the request frequency information included in the request frequency information notification received from the request processing server 10 for each target key, and creates a request frequency information table as illustrated in FIG. create. Here, for example, the replication management server 20 divides the request frequency of each key by a predetermined threshold value, and calculates an integer obtained by rounding off the decimal point as the replication target M in the request frequency information table. Further, the replication management server 20 calculates the number N of replications using the above equation (1).

  Thereafter, the replication management server 20 creates a replication management information table by extracting those having the number of replications of 1 or more from the created request frequency information table, and extracts a difference from the replication management table at the time of previous creation (step S302). ). For example, if there is a replication management information table created last time, the replication management server 20 compares the replication management information table created this time with the number of replications (including newly added records). ), And records with a reduced number of replicas (including deleted records) are extracted. In addition, if the replication management information table created at the previous time does not exist, all the records in the replication management information table created this time are records in which the number of replications is increased.

[Effect of the first embodiment]
As described above, in the information processing system 1 according to the first embodiment, when the request processing server 10 receives an access request including a key, the distributed KVS server 50 that stores data corresponding to the key. Is accessed. Then, the request processing server 10 counts the number of accesses that is the number of accepted access requests for each key. Further, when the copy management server 20 acquires the access count for each key from the request processing server 10 and the acquired access count exceeds a predetermined threshold, the number of copies of the key corresponding to the access request is determined. decide. Subsequently, the replication management server 20 creates a copy of the determined number of keys, and the created copy of the key is stored in any one of the plurality of distributed KVS servers 50. Data is assigned to data corresponding to a key whose access count exceeds a threshold.

  Thereby, in the information processing system 1 according to the first embodiment, it is possible to reduce the congestion of the server and effectively use the resources of the entire system. In other words, as illustrated in FIG. 12, in the conventional distributed KVS system, for example, if access concentrates on a specific key “keyY” by the request processing servers 100A to 100C, the distributed KVS server to which the key is assigned. The load concentrates on 50B, congestion occurs in the distributed KVS server, and the service is affected. In addition, server resources cannot be effectively utilized as an entire information processing system.

  In contrast, in the information processing system 1 according to the first embodiment, the replication management server 20 creates as many replicas as the number of replicas determined according to the request frequency of the reference target key, and accesses the request processing server 10. By converting the reference request for the duplicated key by the conversion wrapper 120 into any of the duplicated keys, it is possible to reduce the uneven load of the distributed KVS server 50 resulting from the uneven access frequency of the key in the distributed KVS. This is useful for applications with many reference requests for a specific key.

  In addition, while using an existing distributed KVS system as it is, and without requiring any special function addition to the application, it is possible to reduce the concentration of server load resulting from the deviation of the key request frequency in the distributed KVS. It is possible to effectively use server resources as the entire information processing system. In addition, no special function addition is required for the request processing application 110, and a general-purpose distributed KVS in the market can be used as it is for the distributed KVS client / server.

[Second Embodiment]
In the first embodiment described above, duplicate keys are randomly assigned to each distributed KVS server 50, and the number of duplicate KVS servers 50 to which duplicate keys are assigned is equal to or greater than the replication target M. A case has been described in which the number of replicas N, which is an expected value, is calculated, and key replicas for the number of replicas N are created.

  However, the embodiment is not limited to this. For example, when creating a copy of M keys for the replication target and assigning the created key, In the case of duplication, it may be attempted to assign a duplicate key until the assignment destination distributed KVS servers are not duplicated.

  In the following, as a second embodiment, when duplicating keys for M replication targets and assigning the created keys, when duplicating with the already created duplicate key assignment destination distributed KVS server, The configuration and processing of the replication management server 20 that tries to allocate the replication key until the allocation destination distributed KVS servers are not duplicated will be described. Note that the description of the same configuration and processing as in the first embodiment is omitted.

  The configuration of the replication management server 20A will be described with reference to FIG. FIG. 13 is a block diagram illustrating a configuration of a replication management server according to the second embodiment. As shown in FIG. 13, the replication management server 20A according to the second embodiment is different from the replication management server 20 according to the first embodiment in that it further includes an assignment destination duplication determination unit 22e. Further, the replication management server 20A according to the second embodiment has information stored in the replication management information storage unit 23a and the request frequency information table storage unit 23b as compared with the replication management server 20 according to the first embodiment. Is also different.

  The replication management information storage unit 23a stores, as a request frequency information table, the number of replicated keys for each replicated reference target key, and a distributed KVS server which is a replication key index and a replication key allocation destination server. Information for identifying 50 is stored. For example, as illustrated in FIG. 14, a “key” that is a character string that identifies a copied reference target key, a “number of copies” that indicates the number of copies of the reference target key, and an index of the copy key. The “replication key index” and the “replication key assignment destination server” that is information for identifying the distributed KVS server 50 that is the assignment destination server of the duplication key are stored in association with each other.

  Further, the request frequency information table storage unit 23b stores the number of reference target keys determined by the determination unit 22b described later together with the request frequency for each reference target key. For example, as illustrated in FIG. 15, a “reference target key” that is a character string for identifying a reference target key, a “request frequency” that is the number of times each reference target key is an access request target, And “replication target M” which is the number of copies of the reference target key to be stored in association with each other. Here, in the request frequency information table storage unit 23b according to the second embodiment, the copy number N is not stored, and the value of the copy target M is stored as the copy number of each key.

  The assignment destination duplication determination unit 22e determines whether or not a key for identifying data corresponding to the copy of the key has already been assigned to the distributed KVS server 50 when the copy of the key is assigned to the data by the assignment unit 22d. judge. More specifically, the assignment destination duplication determination unit 22e determines whether the distributed KVS server 50 to which a new duplication key is assigned overlaps with the distributed KVS server 50 to which a duplication source key or an already created duplication key is assigned. Whether or not is determined from a log or the like.

  If the assignment destination duplication determination unit 22e determines that the duplicate key has already been assigned to the distributed KVS server 50, the assignment unit 22d performs the process of assigning the duplicate key again. Specifically, the assigning unit 22d, when the assignment destination duplication determining unit 22e determines that the duplicate key has already been assigned to the distributed KVS server 50, the created duplicate key (for example, the key “ggg_5”) Is added and a new key (for example, “ggg — 6”) added by 1 to the index is added to the distributed KVS server 50.

  Here, the flow of the copy number change process in the information processing system according to the second embodiment will be described with reference to FIG. FIG. 16 is a sequence diagram for explaining the flow of the copy number changing process in the information processing system according to the second embodiment. FIG. 16 shows a process in which the replication management server 20A determines a change in the number of replications based on the request frequency information of each key transmitted from the access conversion wrapper, and adds or deletes a replication key. In addition, the replication management server 20A periodically performs this process in accordance with the interval at which the request processing server 10 transmits the request frequency information.

  As illustrated in FIG. 16, as in the first embodiment, the access conversion wrapper 120 on each request processing server 10 counts the request frequency for each target key included in the reference request transmitted from the request processing application 110. Request frequency count processing is performed (step S401). Then, the access conversion wrapper 120 periodically transmits the request frequency information as a request frequency information notification to the replication management server 20A (for example, every 5 minutes) (step S402).

  Subsequently, in the replication management server 20A according to the second embodiment, when the request frequency information notification is received from the access conversion wrapper 120 on each request processing server 10, the replication number change determination process is performed as follows ( Step S403).

  First, the request frequency information included in the received request frequency information notification is added up for each reference target key, and a request frequency information table as illustrated in FIG. 15 is created. The method for determining the replication target M is the same as that in the first embodiment. However, unlike the first embodiment, the replication management information table as shown in FIG. Create

  When the replication management information table created last time exists, the replication management server 20A compares it with the replication management information table created this time, and records the number of replications (including newly added records) and the number of replications. Each of the records with a decreased number (including deleted records) is extracted. If the replication management information table created at the previous time does not exist, all the records in the replication management information table created this time are records in which the number of replications has increased. The replication management information table created last time is deleted because it is unnecessary after comparing the replication management tables.

  Next, the replication management server 20A reflects the increase and decrease in the number of replications extracted as described above to the distributed KVS server 50 as addition and deletion of replication keys, and updates the replication management information table to the access conversion wrapper. 120 is notified. The process is performed in the order of a replication key addition process, a replication management information update notification, and a replication key deletion process. This is to prevent the access conversion wrapper from accessing the duplicate key before or after addition.

  In the duplication key addition process, the duplication management server 20A first makes a reference request to the duplication source key (step S404), receives a reference response to the reference request (step S405), and receives data (value) corresponding to the key. ) To get. Next, the following processing is repeated for the increased number of copies.

  Here, a case where the predetermined delimiter is “_” and the number of copies of the key “ggg” is increased from 3 to 4 will be described. The replication management server 20A obtains the maximum value of the replication key index of the key “ggg” that has already been created from the replication management information table. If it is 4, the index of the replication key to be added is 1 from that. An addition request for adding 5 to the distributed KVS server 50C using the key “ggg_5” as a new duplicate key is transmitted (step S406), and an addition response to the addition request is received (step S407).

  Then, the replication management server 20A obtains information on which distributed KVS server 50 the key “ggg — 5” is assigned from the log of the distributed KVS client library. As a result, when the assigned server is the distributed KVS server 50C, it can be seen from the replication management information table that the distributed KVS server 50C has already been assigned the dug key “ggg_2” of which the index is 2. The created replication key ggg_5 transmits a deletion request to be deleted to the distributed KVS server 50C (step S408), and receives a deletion response to the deletion request (step S409).

  Then, the replication management server 20A tries to add a new replication key “ggg_6” obtained by adding 1 to the index (steps S410 and S411). Similarly, the allocation destination server of ggg_6 is acquired from the log of the distributed KVS client library, and it is confirmed that it does not overlap with the already created replication key allocation destination server.

  In this way, the duplication key assignment destination server is determined, and if it is duplicated with the existing duplication key assignment destination server, all duplication key assignment destination servers are created by re-creating the duplication key. Can be avoided.

  After performing the replication key addition process for all the keys whose number of replications has increased as described above, the replication management server 20A issues a replication management information update notification to the access conversion wrapper 120 of each request processing server (step In step S412), a copy management information update response is received as a response (step S413). The replication management information update notification may store all the contents of the new replication management information table, store only the update difference, and the access conversion wrapper 120 creates a new replication management information table to which the update difference is applied. It's okay.

  Although not described in the description of FIG. 16, after the replication management information update notification has been sent to all the request processing servers 10, the replication management server 20A deletes the replication key as in the first embodiment. Process. With respect to the deletion process of the duplicate key, the duplicate key is deleted by the necessary number of deletions from the largest index. Further, the allocation destination server information of each replication key held in the replication management information table may be inconsistent when a distributed KVS server is added or removed. For this reason, when it is detected from the log of the distributed KVS client library that an increase or a decrease of the distributed KVS server 50 has occurred, it is necessary to create all the duplicate keys again.

  As described above, in the information processing system according to the second embodiment, in addition to the effects of the first embodiment, the number of necessary duplicate keys is reduced, so that the storage capacity of the distributed KVS server 50 can be reduced. Thus, it is possible to more reliably suppress the load concentration of the distributed KVS server 50.

[System configuration, etc.]
Further, each component of each illustrated apparatus 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. For example, the acquisition unit 22a and the determination unit 22b may be integrated. Further, all or any part of each processing function performed in each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.

  In addition, among the processes described in the present 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.

  In addition, the above-described function deployment to each server is merely an example. For example, the request processing application, the access conversion wrapper, and the distributed KVS client may be operated on independent servers, and the role of the replication management server is requested. It may be superimposed on one of the processing servers.

  In addition, an example in which there is one type of request processing application is shown, and the access conversion wrapper mediates all access to the distributed KVS server from the request processing application, but it is clear that high frequency reference requests are not performed For a maintenance application or the like, the distributed KVS server may be directly accessed without using an access conversion wrapper. In this case, however, the delimiter used for the duplicate key must be a character that is not used in the normal key.

  In the information processing system, the request processing server may have all the functions of the replication management server. Specifically, the request processing server may have functions corresponding to the acquisition unit 22a, the determination unit 22b, the creation unit 22c, and the allocation unit 22d.

[program]
It is also possible to create a program in which the processing executed by the request processing server 10 described in the above embodiment is described in a language that can be executed by a computer. For example, it is possible to create a key assignment program in which the processing executed by the replication management server 20 according to the first embodiment is described in a language that can be executed by a computer. In this case, when the computer executes the key assignment program, the same effect as in the above embodiment can be obtained. Further, the key assignment program is recorded on a computer-readable recording medium, and the same processing as in the first embodiment is realized by recording the key assignment program on the recording medium and causing the computer to read and execute the key assignment program. May be. Hereinafter, an example of a computer that executes a key assignment program that realizes the same function as that of the replication management server 20 illustrated in FIG. 6 will be described.

  FIG. 17 is a diagram illustrating a computer 1000 that executes a key assignment program. As illustrated in FIG. 17, the computer 1000 includes, for example, a memory 1010, a CPU 1020, a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. These units are connected by a bus 1080.

  The memory 1010 includes a ROM (Read Only Memory) 1011 and a RAM 1012 as illustrated in FIG. The ROM 1011 stores a boot program such as BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1090 as illustrated in FIG. The disk drive interface 1040 is connected to the disk drive 1100 as illustrated in FIG. For example, a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1100. The serial port interface 1050 is connected to, for example, a mouse 1110 and a keyboard 1120 as illustrated in FIG. The video adapter 1060 is connected to a display 1130, for example, as illustrated in FIG.

  Here, as illustrated in FIG. 17, the hard disk drive 1090 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the key assignment program is stored in, for example, the hard disk drive 1090 as a program module in which a command executed by the computer 1000 is described.

  In addition, various data described in the above embodiment is stored as program data in, for example, the memory 1010 or the hard disk drive 1090. Then, the CPU 1020 reads the program module 1093 and the program data 1094 stored in the memory 1010 and the hard disk drive 1090 to the RAM 1012 as necessary, and executes various processing procedures.

  The program module 1093 and the program data 1094 related to the key assignment program are not limited to being stored in the hard disk drive 1090, but are stored in, for example, a removable storage medium and read out by the CPU 1020 via the disk drive or the like. Also good. Alternatively, the program module 1093 and the program data 1094 related to the key assignment program are stored in another computer connected via a network (LAN (Local Area Network), WAN (Wide Area Network), etc.), and the network interface 1070 is stored. Via the CPU 1020.

1 Information processing system 10, 10A to 10E Request processing server 11, 21 Communication processing unit 12, 22 Control unit 12a Access unit 12b Count unit 13, 23 Storage unit 13a Request frequency storage unit 13b Replication management information table storage unit 20, 20A Replication Management server 22a Acquisition unit 22b Determination unit 22c Creation unit 22d Allocation unit 22e Allocation destination duplication determination unit 23 Storage unit 23a Replication management information storage unit 23b Request frequency information table storage unit 30, 30A-30C Client 40 Load distribution device 50, 50A- 50E Distributed KVS server 60, 60A-60C IP network

Claims (4)

A plurality of storage devices that store data and a key that identifies each data in association with each other, a server device that accesses the plurality of storage devices, and a key that is assigned to the data stored in the storage device In an information processing system having a management device,
The server device is
When an access request including the key is received, an access unit that accesses a storage device that stores data corresponding to the key;
For each key, a counting unit that counts the number of accesses that is the number of received access requests;
With
The management device is
An acquisition unit for acquiring the access count for each key from the server device;
When the number of accesses acquired by the acquisition unit exceeds a predetermined threshold , the replication target that is a value obtained by dividing the number of accesses by the threshold is set as M, as the number of copies of the key corresponding to the access request . When the number of storage devices is S, a determination unit that determines the number of copies, which is a value obtained by rounding up the number after the decimal point with respect to the value N calculated by the equation (1) , and
A creation unit that creates a copy of the number of keys determined by the determination unit;
A duplicate of the key created by the creating unit is assigned to data stored in any one of the plurality of storage devices and corresponding to a key whose number of accesses exceeds the threshold. An allocator;
An information processing system comprising:

When the access unit receives an access request including the key, the access unit determines whether the key is a duplicated key. If the key is a duplicated key, the access unit selects one of the duplicated keys. the information processing system according to claim 1, select key, and performs access to the storage device by using the key. An acquisition unit that acquires the number of accesses for each key from a server device that accesses a plurality of storage devices that store data and a key that identifies each data in association with each other;
When the number of accesses acquired by the acquisition unit exceeds a predetermined threshold , the replication target that is a value obtained by dividing the number of accesses by the threshold is set as M, as the number of copies of the key corresponding to the access request . When the number of storage devices is S, a determination unit that determines the number of copies, which is a value obtained by rounding up the number after the decimal point with respect to the value N calculated by the equation (1) , and
A creation unit that creates a copy of the number of keys determined by the determination unit;
A duplicate of the key created by the creating unit is assigned to data stored in any one of the plurality of storage devices and corresponding to a key whose number of accesses exceeds the threshold. An allocator;
A management apparatus comprising:

An acquisition step of acquiring the number of accesses for each key from a server device that accesses a plurality of storage devices that store data and a key that identifies each data in association with each other;
When the number of accesses acquired by the acquisition step exceeds a predetermined threshold , the replication target that is a value obtained by dividing the number of accesses by the threshold is M, as the number of copies of the key corresponding to the access request . A determination step of determining the number of replicas, which is a value obtained by rounding up the number after the decimal point with respect to the value N calculated by the equation (1) when the number of storage devices is S ;
Creating a copy of the number of keys determined by the determining step;
A copy of the key created in the creating step is assigned to data stored in any one of the plurality of storage devices and corresponding to a key whose number of accesses exceeds the threshold. An assignment step;
A key assignment program for causing a computer to execute the above.

Images