JP2009217769A - Resource excess distribution prevention system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To operate a system while stabilizing the whole system. <P>SOLUTION: A resource distribution device 60 includes a transfer object exclusion file 63 in which information about service which is not changed for a predetermined period is stored. Values representing resource states of each of a plurality of servers 1-9, or each of a plurality of services A, B, C are obtained, and a first service A whose value exceeds a first setting value is selected among the plurality of services A, B, C. A second service B except the service which is not changed for the predetermined period and except the first service A is selected among the plurality of services A, B, C by referring to the transfer object exclusion file 63. At least one server 4 among the servers 4-6 assigned to the second service B is assigned to the first service A instead of the second service B by referring to a service assignment file 33. Information considering the first service A as the service which is not changed for the predetermined period is stored in the transfer object exclusion file 63. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a resource overdistribution prevention system that allocates resources.

  A resource distribution system has been developed that allocates a server that provides another service to a resource-deficient service when the resource of a server that provides a service is insufficient.

  For example, it is assumed that there is a server a that provides a certain service A and a server b that provides a service B, and the service A falls short of resources due to access concentration or the like. In the resource distribution system, when it is determined that the resource of the service B has sufficient capacity, the resource is moved from the server b to the server a. Thereby, the resource shortage of the server a is solved.

  At this time, the server b may run out of resources for the service B. In this case, resource distribution from the server a to the server b is performed this time, and the resource shortage of the server b is solved. However, since the state returns to the original state, the server a again becomes in a state of insufficient resources for the service A.

  In this case, server distribution is performed endlessly until the resources of the two servers a and b are recovered simultaneously, and the entire service becomes unstable. It is desirable to operate while stabilizing the entire system.

  Introduces technologies related to resource allocation.

  Japanese Patent Laid-Open No. 2005-141605 describes a computer resource allocation method (Patent Document 1). In the computer resource allocation method, in a computer system in which a plurality of computer resources are connected to a single or a plurality of policy servers, each computer resource is allocated to a plurality of services and operated as an active computer resource. The first group of computer resources waits in the first standby system state in which the respective applications are installed in preparation for the execution of each of the plurality of services, and the second group of remaining computer resources performs the execution of the service. It is characterized by making it wait in the 2nd standby state in which the application for installation is not installed.

  Japanese Patent Application Laid-Open No. 2001-209627 describes an information processing system (Patent Document 2). The information processing system includes a plurality of host systems that operate independently of each other and have sufficient resources including at least an instruction processor and a memory segment. Each host system is physically independent from other components in the host system, and has a system state monitoring change instruction mechanism having a resource operation state management table. It is characterized in that it has a function that enables automatic increase and decrease, and a function that can dynamically increase the capacity by using resources of other host systems if necessary.

  Japanese Patent Application Laid-Open No. 2002-150000 describes a server system (Patent Document 3). The server system responds to a plurality of types of service requests and provides services corresponding to the service requests. This server system is based on at least one server, assignment means for determining an assignment state of a service to be provided for each time period in at least one server, a reception means for receiving a service request from a client, and the assignment state. When it is determined that the service request received by the receiving means is for a service to be provided in the time zone to which the current time belongs, a service request is sent to the service providing module that performs service providing processing for the service request. Delivery means for delivering. The allocating means is characterized by changing an allocating state for at least one server for each service according to each time zone.

JP 2005-141605 A (Claim 1) JP 2001-209627 A (Claim 1) JP 2002-150000 A (Claim 1)

  An object of the present invention is to provide a resource overdistribution prevention system capable of operating a system while stabilizing the entire system.

  In the following, means for solving the problems will be described using the reference numerals used in the best modes and embodiments for carrying out the invention in parentheses. This reference numeral is added to clarify the correspondence between the description of the claims and the description of the best mode for carrying out the invention / example, and is described in the claims. It should not be used to interpret the technical scope of the invention.

The resource overdistribution prevention system of the present invention comprises a service providing system (40), a user device (50) for transmitting a service request to the service providing system (40), and a resource distribution device (60). Yes.
The service providing system (40) includes a plurality of servers (1 to 9), a service allocation file (33), and a load distribution device (30).
The plurality of servers (1 to 9) provide service data corresponding to the service request among the plurality of services (A, B, C) to the user device (50).
The service allocation file (33) includes one of a plurality of services (A, B, C) and one of the servers (1 to 3) of the plurality of servers (1 to 9). 4-6 and 7-9) are stored.
The load balancer (30) refers to the service assignment file (33) and transmits a service request from the user device (50) to a server associated with the corresponding service.
The resource distribution device (60) includes a migration target exclusion file (63), a resource status monitoring unit (64), and a redistribution unit (65).
The migration target exclusion file (63) stores information on services that do not change for a predetermined period.
The resource status monitoring unit (64) acquires a value representing the resource status of each of the plurality of servers (1 to 9) or each of the plurality of services (A, B, C) (S1) (S11) The first service (A) whose value exceeds the first set value (85%) (8.5) is selected from the plurality of services (A, B, C) (S4-YES) (S13-YES) ).
The redistribution unit (65) refers to the movement target exclusion file (63), and selects the first service (A) from a plurality of services (A, B, C) other than a service that does not change for a predetermined period. The second service (B) other than is selected (S5 to S7) (S14 to S17).
The redistribution unit (65) refers to the service allocation file (33) and assigns at least one server (4) among the servers (4 to 6) allocated to the second service (B) to the second Instead of the service (B), the first service (A) is assigned (S7) (S17).
The redistribution unit (65) stores, in the movement target exclusion file (63), information indicating that the first service (A) is not changed for a predetermined period of time (S8) (S18).

  As described above, in the resource overdistribution prevention system of the present invention, even if a resource in a specific service is insufficient, server allocation may be performed from a service that has just recovered from the resource shortage (a service that does not change for a predetermined period). Disappear. As a result, it is possible to prevent continuous server allocation between specific services, and to operate the system while stabilizing the entire system.

  Hereinafter, a resource over-distribution prevention system according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 illustrates a resource overdistribution prevention system according to an embodiment of the present invention. The resource overdistribution prevention system includes a plurality of user devices 50, a service providing system 40, and a resource distribution device 60.

  Each of the plurality of user devices 50 is connected to the service providing system 40 via a public network (for example, the Internet). The user device 50 is a computer, and includes an execution unit 51 that is a CPU (Central Processing Unit), and a storage unit 52 that is a memory in which a computer program to be executed by the execution unit 51 is stored. The execution unit 51 reads a computer program from the storage unit 52 and executes it when the computer is started up. The execution unit 51 transmits a service request for requesting one of the services A, B, C, etc. to the service providing system 40, and receives service data corresponding to the service request.

  The service providing system 40 includes a plurality of servers 1 to 9, a NAS (Network Attached Storage) 20 that is a storage device, and a load distribution device 30. The plurality of servers 1 to 9, the NAS 20, the load balancer 30, and the resource balancer 60 are connected to a network (for example, a LAN (Local Area Network)). The load balancer 30 is connected to a public line network (Internet).

  The NAS 20 has a plurality of storage areas 21 to 23. The plurality of storage areas 21 to 23 store data related to a plurality of services A, B, and C, respectively.

  Each of the plurality of servers 1 to 9 is a computer, and includes an execution unit 13 that is a CPU and a storage unit 14 that is a memory in which a computer program to be executed by the execution unit 13 is stored. The execution unit 13 reads a computer program from the storage unit 14 and executes it when the computer is started up. The computer program includes a service providing unit 15. The service providing unit 15 acquires service data corresponding to the service request from the NAS 20 and transmits the data to the user device 50 that requested the service via the load balancer 30 and the public line network. The servers 1 to 9 may be provided with spare servers that do not necessarily provide each service. The servers 1 to 9 may be connected to a public line network (Internet). The NAS 20 may be incorporated in the servers 1 to 9 as a storage device. In addition, the service providing unit 15 of the servers 1 to 9 transmits the service data corresponding to the service request to the user device 50 that requested the service via the public line network, not via the load balancer 30. May be. Further, the service providing unit 15 of the servers 1 to 9 may perform calculation according to the service request, and transmit the calculation result as service data to the user device 50 that requested the service. In this case, since the service providing unit 15 does not need to refer to the NAS 20, the NAS 20 is not an essential component.

  The load balancer 30 (referred to as a load balancer or the like) is a computer, and includes an execution unit 31 that is a CPU, and a storage unit 32 that is a memory in which a computer program to be executed by the execution unit 31 is stored. Yes. The storage unit 32 includes a service assignment file 33.

  As shown in FIG. 2, the service assignment file 33 includes one of a plurality of services A, B, and C and one of a plurality of servers 1 to 9. Stores information to be assigned. As the above information, it is assumed that servers 1 to 3, 4 to 6, and 7 to 9 are assigned to a plurality of services A, B, and C, respectively.

  The execution unit 31 reads a computer program from the storage unit 32 and executes it when the computer is started. The computer program includes a service management unit 34. When the service management unit 34 receives a service request from the user device 50, the service management unit 34 refers to the service allocation file 33. At this time, the service management unit 34 transmits a service request from the user device 50 to a server (in this case, one of the servers 1 to 3) associated with the corresponding service (for example, service A). In this way, the loads on the plurality of servers 1 to 9 are distributed.

  The resource distribution device 60 is a computer and includes an execution unit 61 that is a CPU and a storage unit 62 that is a memory in which a computer program to be executed by the execution unit 61 is stored. The storage unit 62 includes a movement target exclusion file 63.

  The migration target exclusion file 63 stores information on services that are not changed for a predetermined period. This information represents that a service to which a server is additionally allocated is excluded from being a server allocation source for a predetermined period. In the present embodiment, the movement target exclusion file 63 stores information associating a plurality of services A, B, and C with count values as service information that is not changed for a predetermined period, as shown in FIG. . Alternatively, information on the history of services to which servers are additionally assigned may be stored.

  The execution unit 61 reads a computer program from the storage unit 62 and executes it when the computer is started up. The computer program includes a resource status monitoring unit 64 and a redistribution unit 65. The resource status monitoring unit 64 monitors the plurality of servers 1 to 9 to acquire a value representing the resource status, and the redistribution unit 65 determines the service based on the value and the count value of the migration target exclusion file 63. The server service allocation is changed for the allocation file 33. Details of this will be described later.

  As values representing the resource status, the CPU usage rate, the memory usage rate {obtained directly from the server by the resource distribution device 60 using SNMP (Simple Network Management Protocol)}, and the server response time {TAT (Turn Around Time) to the service request And the resource distribution device 60 obtains from the server via the load balancer 30}.

(First embodiment)
In the resource overdistribution prevention system according to the first embodiment of the present invention, the description overlapping the above-described embodiment is omitted.

  FIG. 4 is a flowchart showing the operation of the resource distribution device 60 in the resource overdistribution prevention system according to the first embodiment of the present invention.

  The resource status monitoring unit 64 monitors the plurality of servers 1 to 9 at regular intervals, and acquires the server resource status value as a value representing the resource status of the plurality of servers 1 to 9 (step S1). The server resource status value includes the above-described CPU utilization rate, memory utilization rate, and the like.

  The resource status monitoring unit 64 uses the average value of the server resource status values of the servers 1 to 3, 4 to 6, and 7 to 9 as the average server resource status value as the resource status values of the plurality of services A, B, and C. (Step S2). Here, although the average value was calculated | required, you may use other values, such as an addition value and the highest value, as a calculated value.

  At this time, the resource status monitoring unit 64 refers to the migration target exclusion file 63 and updates the count values corresponding to the plurality of services A, B, and C. In this case, 1 is subtracted from the count values corresponding to the plurality of services A, B, and C (step S3).

  The resource status monitoring unit 64 selects a first service (for example, service A) whose average server resource status value exceeds the first set value from among the plurality of services A, B, and C (step S4-YES).

  The redistribution unit 65 excludes the first service (service A) from the plurality of services A, B, and C, and the first candidate service (as a candidate service whose average server resource status value is lower than the second set value) For example, services B and C) are selected (step S5). The second set value is lower than the first set value. Here, since the second set value is lower than the first set value, the first service (service A) may not be excluded.

At this time, the redistribution unit 65 refers to the migration target exclusion file 63, and excludes the service whose count value is 1 or more from the first candidate services (services B and C) as a service that does not change for a predetermined period. The selected second candidate service (for example, services B and C) is selected (step S6).
(Also, if the migration target exclusion file 63 stores information on the history of the service to which the server is additionally allocated, it is additionally recorded within the predetermined period (compared with the current date and time and the current period). (Services to which servers are assigned are not subject to movement.)

  The redistribution unit 65 selects at least one service among the second candidate services (services B and C) as the second service (for example, service B). Here, the redistribution unit 65 selects a service having the lowest average server resource status value as the second service (service B) from the second candidate services (services B and C). In this way, the second service (service B) other than the first service (service A) other than the service that does not change for a predetermined period is selected from the plurality of services A, B, and C. Next, the redistribution unit 65 refers to the service assignment file 33, and at least one server (for example, server 4) of servers (in this case, servers 4 to 6) assigned to the second service (service B). 4) is assigned to the first service (service A) instead of the second service (service B) (step S7).

The redistribution unit 65 moves information that associates the first service (service A) with a set count value of 2 or more as a count value in order to make the first service (service A) a service that does not change for a predetermined period. It stores in the exclusion file 63 (step S8).
(In addition, if the movement target exclusion file 63 stores information on the history of services to which servers are additionally assigned, the servers can be additionally assigned in association with the date and number of the period. (If there is a service number assigned in some period, there may be no service to which a server is additionally allocated) in the migration target exclusion file 63.)

  When there is no first candidate service or second candidate service (step S5-no service that can be extracted) (step S6-no service to be moved), the redistribution unit 65 assigns a spare server to the first service (service A). (Step S9) The alarm information indicating the alarm is notified to the computer used by the system administrator by e-mail or the like (Step S10).

[Example]
The first embodiment will be specifically described with reference to FIG. 5, taking as an example the case where the server resource status value is a CPU utilization rate.

Here, the first set value (hereinafter referred to as a tight value) is set to 85%, and the second set value (hereinafter referred to as a surplus value) is set to 55%.
The set count value described above is set to 3.
In addition, a plurality of services A, B, and C, and servers 1 to 3, 4 to 6, and 7 to 9 that provide data of the plurality of services A, B, and C are allocated to the service allocation file 33. Information is stored.
The movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with 0, 0, and 0 as count values.

The resource status monitoring unit 64 transmits a resource status request to each server and receives the resource status of each server at regular intervals (step S1).
The CPU utilization rates in the servers 1 to 3, 4 to 6, and 7 to 9 are 90%, 40%, and 50%, respectively.

The resource status monitoring unit 64 refers to the resource status and the service allocation file 33 of the load balancer 30, obtains the average value of the server resources allocated for each service, and determines the resource tightness / surplus status of each service. Obtained (step S2).
(For simplicity of explanation, the resource is set as the CPU usage rate and the average value is obtained. However, the calculation may use other values such as an addition value and the highest value. In addition, the resource tightness / surplus status of each service is obtained by the load balancer 30, and the resource balancer acquires the resource tightness / surplus status of each service from the load balancer 30. It does not matter.)

The resource status monitoring unit 64 updates the count value in the migration target exclusion file 63 (step S3).
(This process may be performed anywhere in the procedure, and may be performed by any part. It may be performed at regular intervals or when the server allocation is changed. In FIG. The value is decreased by 1. The initial value of each service in the migration target exclusion file 63 is 0, and it is assumed that it is not 0 or less.)

  The resource status monitoring unit 64 extracts a service in which the resource tightness / surplus status of each service exceeds a predetermined tightness value (step S4-YES).

When the service is extracted, the redistribution unit 65 identifies a service that is lower than the predetermined surplus value from the resource tightness / surplus status of each service (step S5).
(In FIG. 5, the service A is extracted when it exceeds the tight value, and the service B and service C are extracted when it is below the surplus value.)

The redistribution unit 65 refers to the movement target exclusion file 63, determines that the extracted service is not a movement target exclusion when it falls below the surplus value, and excludes the service that is not the movement target (step S6).
(In FIG. 5, if the count value of each service in the migration target exclusion file 63 is 1 or more, the service is excluded from the migration target. In FIG. 5, each service 0 of the service B and service C is a migration target.)
(Also, if the migration target exclusion file 63 stores information on the history of the service to which the server is additionally allocated, it is additionally recorded within the predetermined period (compared with the current date and time and the current period). (Services to which servers are assigned are not subject to movement.)

From the services that are not excluded from the movement target, the service of the resource of the lowest service is identified as a tight / surplus service. (In FIG. 5, service B.)
With reference to the service allocation file 33, one of the servers allocated to the lowest service is allocated and changed to a service exceeding the tightness value (step S7).
(In FIG. 5, the server 4 is changed from the server 4, the server 5, and the server 6 assigned to the service B to the service A by rewriting and changing the service assignment file 33 of the load balancer 30. Although there is one service or server for changing the provided service, there are cases where a plurality of servers or a plurality of services are changed instead of one (for example, one server can be changed from a plurality of surplus situations). Yes, you can set the number of server changes in the configuration file.)

The count value in the migration target exclusion file 63 of the service to which the server is assigned is excluded from the migration target (step S8).
(In FIG. 5, the count value in the service A is set to 3. That is, the service A is excluded from the movement target in the two periods in which the resource status is acquired. The number of values is arbitrary.)
(In addition, if the movement target exclusion file 63 stores information on the history of services to which servers are additionally assigned, the servers can be additionally assigned in association with the date and number of the period. (If there is a service number assigned in some period, there may be no service to which a server is additionally allocated) in the migration target exclusion file 63.)

  When there is no service that is less than the surplus value or service that is not excluded from the movement target (step S5-no service that can be extracted) (step S6-no movement target service), the spare server is set as a tight service (step S9), Alarm information is transmitted to the system administrator by e-mail or the like (step S10).

  Hereinafter, the above continuation will be described with reference to FIG.

Here, the service allocation file 33 includes a plurality of services A, B, and C, and servers 1 to 4, 5 to 6, and 7 to 9 that provide data of the plurality of services A, B, and C, respectively. Stores information to be assigned.
The movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 3, 0, and 0.

After a certain period, the resource distribution device 60 receives the resource status of each server and similarly obtains the resource status for each service.
The CPU utilization rates in the servers 1 to 4, 5 to 6, and 7 to 9 are 68%, 60%, and 50%, respectively. In this case, the resource tightness / surplus state of each service is calculated as 68%, 60%, and 50%.
The count value for each service in the migration target exclusion file 63 is subtracted by 1.
As a result, the movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 2, 0, and 0.
This time, since there is no service that exceeds 85% of the tight value, it ends.

  Here, the service allocation file 33 includes a plurality of services A, B, and C, and servers 1 to 4, 5 to 6, and 7 to 9 that provide data of the plurality of services A, B, and C, respectively. Stores information to be assigned.

After a certain period, the resource distribution device 60 receives the resource status of each server and similarly obtains the resource status for each service.
The CPU utilization rates in the servers 1 to 4, 5 to 6, and 7 to 9 are 45%, 90%, and 50%, respectively. In this case, the resource tightness / surplus state of each service is calculated as 45%, 90%, and 50%.
The count value for each service in the migration target exclusion file 63 is subtracted by 1.
As a result, the movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 1, 0, and 0.
This time, service B is extracted as a service exceeding the constraining value of 85%, and service A and service C are extracted as services falling below the surplus value of 55%. With reference to the movement target exclusion file 63, since the service A is 1 or more, it is excluded from the movement target, and the service C becomes the movement target. The server of service C is identified as server 7, server 8, and server 9 by referring to the service assignment file 33. The service allocation file 33 is changed so that the server 7 can be allocated to the service B. Further, the count value in the service C of the migration target exclusion file 63 is set to 3.

Here, the service allocation file 33 includes a plurality of services A, B, and C, and servers 1 to 4, 5 to 7, and 8 to 9 that provide data of the plurality of services A, B, and C, respectively. Stores information to be assigned.
The movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 1, 0, and 3.

After a certain period, the resource distribution device 60 receives the resource status of each server and similarly obtains the resource status for each service.
The CPU utilization rates in the servers 1 to 4, 5 to 7, and 8 to 9 are 45%, 60%, and 75%, respectively. In this case, the resource tightness / surplus state of each service is calculated as 45%, 60%, and 75%.
The count value for each service in the migration target exclusion file 63 is subtracted by 1.
As a result, the movement target exclusion file 63 stores information associating a plurality of services A, B, and C with 0, 0, and 2 as count values.
This time, since there is no service that exceeds 85% of the tight value, it ends.

[effect]
As described above, in the resource overdistribution prevention system according to the first embodiment of the present invention, if a resource in a specific service is insufficient, a service that has just recovered from the resource shortage (a service that does not change for a predetermined period, ie, a count) Server allocation is not performed from a service having a positive value. As a result, it is possible to prevent continuous server allocation between specific services, and to operate the system while stabilizing the entire system.

  In the resource overdistribution prevention system according to the first embodiment of the present invention, the server allocation is prevented from being continuously performed among specific services by the method of counting down the count value in the migration target exclusion file 63. It is not limited. As a variation thereof, the server allocation between specific services can also be prevented by the method of counting up the count value in the migration target exclusion file 63.

That is, in step S3, the resource status monitoring unit 64 refers to the migration target exclusion file 63 and updates the count values corresponding to the plurality of services A, B, and C. In this case, 1 is added to the count values corresponding to the plurality of services A, B, and C.
In step S <b> 6, the redistribution unit 65 refers to the movement target exclusion file 63, and the count value is the set count value “ A second candidate service (for example, services B and C) excluding services less than 3 ″ is selected.
In step S8, the redistribution unit 65 excludes information associating the first service (service A) with 0 as the count value so that the first service (service A) does not change for a predetermined period. Store in file 63.

(Second Embodiment)
In the resource overdistribution prevention system according to the second exemplary embodiment of the present invention, the description overlapping with the above exemplary embodiment is omitted.

  FIG. 6 is a flowchart showing the operation of the resource distribution device 60 in the resource overdistribution prevention system according to the second embodiment of the present invention.

  The resource status monitoring unit 64 monitors the plurality of services A, B, and C at regular intervals, and acquires service resource status values as values representing the resource status of the plurality of services A, B, and C (step S11). . The service resource status value includes the TAT described above.

  At this time, the resource status monitoring unit 64 refers to the migration target exclusion file 63 and updates the count values corresponding to the plurality of services A, B, and C. In this case, 1 is subtracted from the count values corresponding to the plurality of services A, B, and C (step S12).

  The resource status monitoring unit 64 selects a first service (for example, service A) whose service resource status value exceeds the first set value from among the plurality of services A, B, and C (step S13—YES).

  The redistribution unit 65 excludes the first service (service A) from the plurality of services A, B, and C, and the first candidate service (for example, as a candidate service whose service resource status value is lower than the second set value) Service B, C) is selected (step S14). The second set value is lower than the first set value. Here, since the second set value is lower than the first set value, the first service (service A) may not be excluded.

At this time, the redistribution unit 65 refers to the migration target exclusion file 63, and excludes the service whose count value is 1 or more from the first candidate services (services B and C) as a service that does not change for a predetermined period. The selected second candidate service (for example, services B and C) is selected (step S15).
(Also, if the migration target exclusion file 63 stores information on the history of the service to which the server is additionally allocated, it is additionally recorded within the predetermined period (compared with the current date and time and the current period). (Services to which servers are assigned are not subject to movement.)

  The redistribution unit 65 monitors the plurality of servers 1 to 9 at regular intervals, and acquires server resource status values representing resources of the plurality of servers 1 to 9 (step S16). Alternatively, the redistribution unit 65 may acquire the server resource status value by issuing a resource status request to the plurality of servers 1 to 9 and receiving the resource status of the servers 1 to 9.

  The redistribution unit 65 selects at least one service among the second candidate services (services B and C) as the second service (for example, service B). Here, the redistribution unit 65 selects the second service (service B) having the lowest server resource status value from the second candidate services (services B and C). In this way, the second service (service B) other than the first service (service A) other than the service that does not change for a predetermined period is selected from the plurality of services A, B, and C. Next, the redistribution unit 65 refers to the service assignment file 33, and at least one server (for example, server 4) of servers (in this case, servers 4 to 6) assigned to the second service (service B). 4) is assigned to the first service (service A) instead of the second service (service B) (step S17).

  Here, the selection of the second service (service B) in step S17 will be described in detail. First, the redistribution unit 65 selects a surplus server whose server resource status value is equal to or less than the third set value from among the servers 4 to 6 and 7 to 9 allocated to the second candidate service (services B and C). select. Next, the service having the largest number of surplus servers is selected as the second service (service B) from the second candidate services (services B and C).

  Further, another method for selecting the second service (service B) will be described in detail. The redistribution unit 65 calculates the average value of the server resource status values in the servers 4 to 6 and 7 to 9 allocated to the second candidate service (services B and C) as the average server resource status value. Here, although the average value was calculated | required, you may use other values, such as an addition value and the highest value, as a calculated value. Next, the service with the lowest average server resource status value is selected as the second service (service B) from the second candidate services (services B and C).

The redistribution unit 65 moves information that associates the first service (service A) with a set count value of 2 or more as a count value in order to make the first service (service A) a service that does not change for a predetermined period. It stores in the exclusion file 63 (step S18).
(In addition, if the movement target exclusion file 63 stores information on the history of services to which servers are additionally assigned, the servers can be additionally assigned in association with the date and number of the period. (If there is a service number assigned in some period, there may be no service to which a server is additionally allocated) in the migration target exclusion file 63.)

  When there is no first candidate service or second candidate service (step S14—no service to be extracted) (step S15—no movement target service), the redistribution unit 65 assigns a spare server to the first service (service A). (Step S19) The alarm information representing the alarm is notified to the computer used by the system administrator by e-mail or the like (Step S20).

[Example]
The second embodiment will be specifically described with reference to FIG. 7, taking as an example the case where the service resource status value is TAT. In FIG. 7, the value of TAT (unit: second or millisecond) is expressed by 1 to 10 for the sake of simplicity of explanation.

Here, the first set value (hereinafter referred to as a tight value) is set to 8.5, and the second set value (hereinafter referred to as a surplus value) is set to 5.5.
The set count value described above is set to 3.
In addition, a plurality of services A, B, and C, and servers 1 to 3, 4 to 6, and 7 to 9 that provide data of the plurality of services A, B, and C are allocated to the service allocation file 33. Information is stored.
The movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with 0, 0, and 0 as count values.
Further, it is assumed that the server resource status value is a CPU utilization rate.
Further, the above-mentioned third set value (hereinafter referred to as the second surplus value) is set to 50%.

The resource status monitoring unit 64 transmits service requests for the services A, B, and C via the load balancer 30 at regular intervals, receives responses, and acquires (measures) the resource status (TAT) of each service. (Step S11).
The TATs for services A, B, and C are 9.0, 4.0, and 5.0, respectively.
Here, by obtaining the resource status several times and obtaining an average value, the resource status can be grasped more accurately.
(TAT actually sees the same effect as a service request from the user device 50, and therefore via the load balancer 30. In the case of TAT, the resource status of the service can be grasped directly, not each server. CPU Since the usage rate of memory and memory is taken as an instantaneous value, it is difficult to grasp the influence on the user personally.)

The resource status monitoring unit 64 updates the count value in the migration target exclusion file 63 (step S12).
(This process may be performed anywhere in the procedure, and may be performed by any part. It may be performed at regular time intervals or when the server allocation is changed. In FIG. The value is decreased by 1. The initial value of each service in the migration target exclusion file 63 is 0, and it is assumed that it is not 0 or less.)

  The resource status monitoring unit 64 extracts a service in which the resource tightness / surplus status of each service exceeds a predetermined tightness value (YES in step S13).

When the service is extracted, the redistribution unit 65 identifies a service that is lower than the predetermined surplus value from the resource status of each service (step S14).
(In FIG. 7, the service A is extracted when it exceeds the tight value, and the service B and service C are extracted when it is below the surplus value.)

The redistribution unit 65 refers to the movement target exclusion file 63, determines whether the extracted service is not a movement target exclusion when the surplus value falls below the surplus value, and excludes the service that is not the movement target (step S15).
(In FIG. 7, if the count value of each service in the migration target exclusion file 63 is 1 or more, the service is excluded from the migration target. In FIG. 7, each service 0 of the service B and service C is a migration target.)
(Also, if the migration target exclusion file 63 stores information on the history of the service to which the server is additionally allocated, it is additionally recorded within the predetermined period (compared with the current date and time and the current period). (Services to which servers are assigned are not subject to movement.)

A resource status request is issued to all servers (services that are less than the surplus value and only the server of the movement target service), and the resource status of each server is received. Here, the resource status is different from the resource status of the service, but it may be the same. (Here, CPU usage rate.)
The CPU usage rates of servers 1, 2, 3, 4, 5, 6, 7, 8, and 9 are 90%, 87%, 89%, 38%, 41%, 39%, 51%, 51%, 49%.
The value of the second resource status of the service, for example, the resource status of the server, the service allocation file 33, and the second surplus value (in this case, 50%), such as the service with the largest percentage of servers with surplus resources. ) To calculate. Among the candidates for the services that are below the surplus value and the services to be moved (service B and service C in FIG. 7), the second resource statuses of the services are compared to obtain a server allocation source service (step S16).
(In FIG. 7, the value of the second resource status is 3 as the number of servers in which the resource status values of the servers 4, 5 and 6 belonging to the service B are equal to or less than the second surplus value 50%. Similarly, the value of the second resource status of the service C is set to 1 with the value of the resource status of the server 7, the server 8, and the server 9 belonging to the service C being the second surplus value of 50% or less. In this case, the server allocation source service is the service B having the highest value (number of surplus servers) of the second resource status of the service C. The value of the second resource status of the service Is the average value of the CPU utilization rate of the servers allocated to each service, and the service with the second resource status value of the lowest service may be used as the server allocation source service.)

The second resource status of each service below the surplus value and a server other than the candidate for the service to be moved (service A in FIG. 7) is more than the tight value when the number of allocation source servers is plural or the tight value. This is used to calculate how many servers are allocated to which service when there are multiple services.
(For example, in FIG. 8, there are two services A and D, which are in a tight state, and there are two services A: 2 and a service D: more than the second tight value: 90%. (Thus, three original servers are required, and servers are allocated to service A and service D at a ratio of 2: 1.)

The server allocated to the server allocation source service is referred to the service allocation file 33, and one of the servers is allocated and changed to a service exceeding the tight value (step S17).
(In FIG. 7, the server 4 is changed from the server 4, the server 5, and the server 6 assigned to the service B to the service A, and the service assignment file 33 of the load balancer 30 is rewritten and changed. Although there is only one service that changes the provided service, there may be cases where multiple servers are changed instead of one (the server can be changed one by one from multiple services in surplus status). Number can be set.)

The count value in the movement target exclusion file 63 of the service to which the server is assigned is excluded from the movement target (step S18).
(In FIG. 7, the count value in the service A is set to 3. That is, the service A is not moved during the two periods for acquiring the resource status. The number of values is arbitrary.)
(In addition, if the movement target exclusion file 63 stores information on the history of services to which servers are additionally assigned, the servers can be additionally assigned in association with the date and number of the period. (If there is a service number assigned in some period, there may be no service to which a server is additionally allocated) in the migration target exclusion file 63.)

  If there is no service that is less than the surplus value or service that is not excluded from the movement target (step S14-no service that can be extracted) (step S15-no movement target service), the spare server is set as a tight service (step S19), The alarm unit transmits alarm information to the system administrator by e-mail or the like (step S20).

  Hereinafter, the continuation will be described with reference to FIG.

Here, the service allocation file 33 includes a plurality of services A, B, and C, and servers 1 to 4, 5 to 6, and 7 to 9 that provide data of the plurality of services A, B, and C, respectively. Stores information to be assigned.
The movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 3, 0, and 0.

In this case, after a certain period of time, the resource distribution device 60 receives a plurality of TATs for each service, and similarly averages to obtain the resource status for each service.
The TATs for services A, B, and C are 6.8, 6.0, and 5.0, respectively.
The count value for each service in the migration target exclusion file 63 is subtracted by 1.
As a result, the movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 2, 0, and 0.
This time, since there is no service exceeding the tight value 8.5.

  Here, the service allocation file 33 includes a plurality of services A, B, and C, and servers 1 to 4, 5 to 6, and 7 to 9 that provide data of the plurality of services A, B, and C, respectively. Stores information to be assigned.

In this case, after a certain period of time, the resource distribution device 60 receives a plurality of TATs for each service, and similarly averages to obtain the resource status for each service.
The TATs for the services A, B, and C are 4.5, 9.0, and 5.0, respectively.
The count value for each service in the migration target exclusion file 63 is subtracted by 1.
As a result, the movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 1, 0, and 0.
This time, service B is extracted as a service exceeding the tight value 8.5, and service A and service C are extracted as services below the surplus value 5.5. With reference to the movement target exclusion file 63, since the service A is 1 or more, it is excluded from the movement target, and the service C becomes the movement target.

Here, when the resource status of each server is received, the CPU utilization rates in the servers 1, 2, 3, 4, 5, 6, 7, 8, 9 are 44%, 45%, 46%, 45%, 91%, 92%, 48%, 50%, 49%, but now there is only one candidate for each service that is below the surplus value and the service to be moved, so the second resource status of the service The calculation of the value is omitted, and the service C becomes the server allocation source service.
In the resource distribution device 60, the server of service C is identified as server 7, server 8, and server 9 by referring to the service allocation file 33. The service allocation file 33 is changed so that one of the servers 7 can be allocated to the service B. Further, the count value in the service C of the migration target exclusion file 63 is set to 3.

Here, the service allocation file 33 includes a plurality of services A, B, and C, and servers 1 to 4, 5 to 7, and 8 to 9 that provide data of the plurality of services A, B, and C, respectively. Stores information to be assigned.
The movement target exclusion file 63 stores information that associates a plurality of services A, B, and C with count values of 1, 0, and 3.

In this case, after a certain period of time, the resource distribution device 60 receives a plurality of TATs for each service, and similarly averages to obtain the resource status for each service.
The TATs for services A, B, and C are 4.5, 6.0, and 7.5, respectively.
The count value for each service in the migration target exclusion file 63 is subtracted by 1.
As a result, the movement target exclusion file 63 stores information associating a plurality of services A, B, and C with 0, 0, and 2 as count values.
This time, since there is no service exceeding the tight value 8.5.

[effect]
As described above, in the resource overdistribution prevention system according to the second embodiment of the present invention, if a resource in a specific service is insufficient, a service that has just recovered from the resource shortage (a service that does not change for a predetermined period, ie, a count) Server allocation is not performed from a service having a positive value. As a result, it is possible to prevent continuous server allocation between specific services, and to operate the system while stabilizing the entire system.

  In the resource overdistribution prevention system according to the second embodiment of the present invention, the server allocation is prevented from being continuously performed between specific services by the method of counting down the count value in the migration target exclusion file 63. It is not limited. As a variation thereof, the server allocation between specific services can also be prevented by the method of counting up the count value in the migration target exclusion file 63.

That is, in step S12, the resource status monitoring unit 64 refers to the migration target exclusion file 63 and updates the count values corresponding to the plurality of services A, B, and C. In this case, 1 is added to the count values corresponding to the plurality of services A, B, and C.
In step S15, the redistribution unit 65 refers to the migration target exclusion file 63 and selects a service whose count value is less than the set count value “3” from the first candidate services (services B and C). The excluded second candidate service (for example, services B and C) is selected.
In step S <b> 18, the redistribution unit 65 stores information that associates the first service (service A) with 0 as the count value in the movement target exclusion file 63.

FIG. 1 illustrates a resource overdistribution prevention system according to an embodiment of the present invention. FIG. 2 shows the service assignment file 33. FIG. 3 shows the migration target exclusion file 63. FIG. 4 is a flowchart showing the operation of the resource distribution device 60 in the resource overdistribution prevention system according to the first embodiment of the present invention. FIG. 5 is a diagram for explaining the operation of the resource distribution device 60 in the resource overdistribution prevention system according to the first embodiment of the present invention. FIG. 6 is a flowchart showing the operation of the resource distribution device 60 in the resource overdistribution prevention system according to the second embodiment of the present invention. FIG. 7 is a diagram for explaining the operation of the resource distribution device 60 in the resource overdistribution prevention system according to the second embodiment of the present invention. FIG. 8 is a diagram for explaining the operation of the resource distribution device 60 in the resource overdistribution prevention system according to the second embodiment of the present invention.

Explanation of symbols

1-9 servers,
10-12 servers (no configuration diagram),
13 execution part,
14 storage section,
15 Service provision department,
20 NAS,
21-23 storage area,
30 load balancer,
31 execution part,
32 storage,
33 Service allocation file,
34 Service Management Department,
40 service provision system,
50 user equipment,
51 execution unit,
52 storage,
60 resource distribution device,
61 execution part,
62 storage,
63 Excluded files to be moved,
64 Resource status monitoring unit,
65 Redistribution Department,

Claims (18)

A service providing system;
A user device for transmitting a service request to the service providing system;
A resource distribution device;
Comprising
The service providing system includes:
Among a plurality of services, a plurality of servers that provide service data corresponding to the service request to the user device;
A service allocation file storing information for allocating any one of the plurality of services and any one of the plurality of servers;
A load distribution device that refers to the service allocation file and transmits the service request from the user device to a server associated with the corresponding service;
With
The resource distribution device includes:
A migration exclusion file that contains information about services that do not change for a specified period of time;
A resource status for acquiring a value indicating a resource status of each of the plurality of servers or each of the plurality of services, and selecting a first service in which the value exceeds a first set value from the plurality of services. A monitoring unit;
A redistribution unit that selects a second service other than the service that does not change for a predetermined period from the plurality of services and refers to the migration target exclusion file;
Comprising
The redistribution unit
Referring to the service assignment file, assign at least one of the servers assigned to the second service to the first service instead of the second service;
Storing information on the first service as a service that does not change for a predetermined period of time in the movement exclusion file;
Resource over-distribution prevention system. The redistribution unit
A first candidate service is selected from the plurality of services, the value being lower than a second setting value lower than the first setting value;
Referring to the movement target exclusion file, select a second candidate service other than the service that does not change for a predetermined period from the first candidate service and the second candidate service other than the first service. Selecting at least one service as a second service;
The resource overdistribution prevention system according to claim 1. The resource status monitoring unit
Obtaining a service resource status value representing the resource status of each of the plurality of services as the value;
Selecting the first service from among the plurality of services, wherein the service resource status value exceeds the first set value;
The redistribution unit
Selecting the first candidate service from among the plurality of services, wherein the service resource status value is lower than the second set value;
With reference to the migration target exclusion file, the second candidate service is selected from the first candidate services,
Obtaining a server resource status value representing each resource of the plurality of servers;
Selecting the second service having the lowest server resource status value from the second candidate services;
The resource overdistribution prevention system according to claim 2. The redistribution unit
From the servers that provide the data of the second candidate service, select a surplus server whose server resource status value is equal to or less than a third setting value;
A service having the largest number of surplus servers is selected as the second service from the second candidate services;
The resource overdistribution prevention system according to claim 3. The resource status monitoring unit
Obtaining a server resource status value representing the resource status of each of the plurality of servers as the value;
Selecting the first service from among the plurality of services, wherein the server resource status value exceeds the first setting value;
The redistribution unit
From the plurality of services, select the first candidate service whose server resource status value is lower than the second setting value;
With reference to the migration target exclusion file, the second candidate service is selected from the first candidate services,
Selecting the second service having the lowest server resource status value from the second candidate services;
The resource overdistribution prevention system according to claim 2. The resource status monitoring unit
As the value of the resource status of each of the plurality of servers, an average value of the server resource status values in a server that provides data of each of the plurality of services is calculated as an average server resource status value,
From the plurality of services, select the first service in which the average server resource status value exceeds the first setting value;
The redistribution unit
Selecting the first candidate service from among the plurality of services, wherein the average server resource status value is lower than the second set value;
Selecting the second service having the lowest average server resource status value from the second candidate services;
The resource overdistribution prevention system according to claim 5. The movement target exclusion file stores information associating the plurality of services with a count value as information on services that do not change for a predetermined period of time.
The resource status monitoring unit
When obtaining the value, referring to the migration target exclusion file, subtract 1 as an update of the count value corresponding to the plurality of services,
The redistribution unit
With reference to the migration target exclusion file, a service that excludes a service having the count value of 1 or more from the plurality of services is selected as the second service,
In order to make the first service a service that does not change for a predetermined period, information associating the first service with the set count value of 2 or more as the count value is stored in the migration target exclusion file.
The resource overdistribution prevention system according to any one of claims 1 to 6. The movement target exclusion file stores information associating the plurality of services with a count value as information on services that do not change for a predetermined period of time.
The resource status monitoring unit
When acquiring the value, referring to the migration target exclusion file, add 1 as the update of the count value corresponding to the plurality of services,
The redistribution unit
With reference to the migration target exclusion file, a service that excludes a service whose count value is less than a set count value of 2 or more is selected as the second service from the plurality of services.
In order to make the first service a service that does not change for a predetermined period, information associating the first service with the count value of 0 is stored in the migration target exclusion file.
The resource overdistribution prevention system according to any one of claims 1 to 6.   A resource distribution device applied to the resource overdistribution prevention system according to claim 1. A service providing system, and a user device that transmits a service request to the service providing system,
The service providing system includes, among a plurality of services, a plurality of servers that provide service data corresponding to the service request to the user device, a service allocation file, and a load distribution device.
The service allocation file stores information for allocating any one of the plurality of services and any one of the plurality of servers,
The load distribution device refers to the service allocation file and transmits the service request from the user device to a server associated with the corresponding service;
A resource distribution method using a computer applied to a resource overdistribution prevention system,
Obtaining a value representing a resource status of each of the plurality of servers or each of the plurality of services;
Selecting a first service in which the value exceeds a first set value from the plurality of services;
A step of selecting a second service other than a service that does not change for a predetermined period and a service other than the first service from among the plurality of services with reference to a migration target exclusion file that stores information on services that do not change for a predetermined period When,
Allocating at least one server among servers allocated to the second service to the first service instead of the second service with reference to the service allocation file;
Storing information on the first service as a service that does not change for a predetermined period in the migration target exclusion file;
A resource distribution method comprising: The step of selecting the second service includes:
A first candidate service is selected from the plurality of services, the value being lower than a second setting value lower than the first setting value;
Referring to the movement target exclusion file, select a second candidate service other than the service that does not change for a predetermined period from the first candidate service and the second candidate service other than the first service. Selecting at least one service as a second service;
The resource distribution method according to claim 10. Obtaining the value comprises:
Obtaining a service resource status value representing the resource status of each of the plurality of services as the value;
The step of selecting the first service includes:
Selecting the first service from among the plurality of services, wherein the service resource status value exceeds the first set value;
The step of selecting the first candidate service includes:
Selecting the first candidate service from among the plurality of services, wherein the service resource status value is lower than the second set value;
The step of selecting the second service includes:
With reference to the migration target exclusion file, the second candidate service is selected from the first candidate services,
Obtaining a server resource status value representing each resource of the plurality of servers;
Selecting the second service having the lowest server resource status value from the second candidate services;
The resource distribution method according to claim 11. The step of selecting the second service includes:
From the servers that provide the data of the second candidate service, select a surplus server whose server resource status value is equal to or less than a third setting value;
A service having the largest number of surplus servers is selected as the second service from the second candidate services;
The resource distribution method according to claim 12. Obtaining the value comprises:
Obtaining a server resource status value representing the resource status of each of the plurality of servers as the value;
The step of selecting the first service includes:
Selecting the first service from among the plurality of services, wherein the server resource status value exceeds the first setting value;
The step of selecting the first candidate service includes:
From the plurality of services, select the first candidate service whose server resource status value is lower than the second setting value;
The step of selecting the second service includes:
With reference to the migration target exclusion file, the second candidate service is selected from the first candidate services,
Selecting the second service having the lowest server resource status value from the second candidate services;
The resource distribution method according to claim 11. Calculating, as the average server resource status value, an average value of the server resource status values in a server that provides data of each of the plurality of services, as the value of the resource status of each of the plurality of servers;
Further comprising
The step of selecting the first service includes:
From the plurality of services, select the first service in which the average server resource status value exceeds the first setting value;
The step of selecting the first candidate service includes:
Selecting the first candidate service from among the plurality of services, wherein the average server resource status value is lower than the second set value;
The step of selecting the second service includes:
Selecting the second service having the lowest average server resource status value from the second candidate services;
The resource distribution method according to claim 14. The movement target exclusion file stores information associating the plurality of services with a count value as information on services that do not change for a predetermined period of time.
The resource distribution method is:
Subtracting 1 as an update of the count value corresponding to the plurality of services with reference to the migration target exclusion file when the value is acquired;
Further comprising
The step of selecting the second service includes:
With reference to the migration target exclusion file, a service that excludes a service having the count value of 1 or more from the plurality of services is selected as the second service,
The step of storing in the movement exclusion file the information that the first service is a service that does not change for a predetermined period of time,
Storing information associating the first service with the set count value of 2 or more as the count value in the migration target exclusion file;
The resource distribution method according to claim 10. The movement target exclusion file stores information associating the plurality of services with a count value as information on services that do not change for a predetermined period of time.
The resource distribution method is:
A step of adding 1 as an update of the count value corresponding to the plurality of services with reference to the migration target exclusion file when the value is acquired;
Further comprising
The step of selecting the second service includes:
With reference to the migration target exclusion file, a service that excludes a service whose count value is less than a set count value of 2 or more is selected as the second service from the plurality of services.
The step of storing in the movement exclusion file the information that the first service is a service that does not change for a predetermined period of time,
Storing information associating the first service with the count value of 0 in the migration target exclusion file;
The resource distribution method according to claim 10.   A computer program that causes the computer to execute each step of the resource distribution method according to claim 10.

Images