JP3987517B2 - Computer system and cluster system program - Google Patents

Description

  The present invention generally relates to a computer system including a plurality of computers, and more particularly to a cluster system technology that realizes an optimal service allocation function according to a computer failure or a load situation.

In recent years, there has been a software technique called a cluster system that manages a computer system composed of a plurality of computers (for example, servers) and improves the processing performance and reliability of services provided to client terminals (users) by executing application programs. Has been developed. The cluster system responds to computer start-up, failure occurrence, and changes in load conditions.
It has functions such as scheduling services running on a computer system to the optimal computer, improving availability and load balancing.

  Cluster systems are roughly classified into a load distribution type cluster system that places importance on the load distribution function and a highly available type cluster system that places importance on the failover function (see, for example, Non-Patent Document 1 and Non-Patent Document 2).

  The cluster system determines an optimal computer for executing the service based on the setting of policy information corresponding to the rules for system operation. Normally, policy information can be changed by user settings.

Further, the cluster system responds by using a spare computer (provisioning computer) when all the computers are in a high load state and there is no optimum computer to which a service is allocated.
R. Buyya, “High Performance Cluster Computing: Architecture and Systems (Volume 1 & 2)”, 1999, Prentice Hall Tetsuo Kaneko, Yoshiya Mori, "Cluster Software", Toshiba Review, Vol.54 No.12 (1999), p.18-21

  In recent years, a cluster system in which a load balancing cluster system and a highly available cluster system are mixed has been developed. In such a system, when the optimal arrangement of services (assignment of services to the optimal computer) is performed simply by setting the policy information, the execution of the service according to the change in the load status of the computer is not guaranteed. Things happen. Specifically, when automatic switchover of services is performed, frequent switchovers occur due to load fluctuations, or the response when the low priority service was executed first is unknown, Also, when there was no computer that could execute the service, it could not be started.

  Accordingly, an object of the present invention is to realize a cluster system capable of reliably performing a service relocation according to a service execution status such as a dynamic load status change after the optimal service allocation. .

  A computer system having two or more computers according to an aspect of the present invention includes a policy management unit that stores policy information for determining allocation processing of a plurality of services executed by each computer, and the policy information A service optimal allocation unit that executes a process of allocating each service to an optimal computer, and a service relocation process allocated by the service optimal allocation unit according to the execution status of the service between the computers. And a service relocation unit executed by referring to the data.

  According to another aspect of the present invention, particularly in a composite cluster system in which a load balancing cluster system and a highly available cluster system coexist, an optimal arrangement of services between cluster systems in response to a change in dynamic load status A computer system having a configuration enabling the above is configured.

  According to the present invention, it is possible to realize a cluster system capable of reliably rearranging services in accordance with service execution status such as dynamic load status change after optimal service allocation.

  Embodiments of the present invention will be described below with reference to the drawings.

(First embodiment)
FIG. 1 is a block diagram illustrating a system configuration of a computer system according to the first embodiment.

  The computer system has a configuration in which, for example, four computers C1 to C4 are connected to each other on the network N. The computers C1 to C4 operate under the control of operating systems (OS-1 to OS-4) 31 to 34, respectively. Here, one spare computer (provisioning computer) C5 is connected to the computer system via the network N. Furthermore, one or more spare computers may be connected to the network N.

  A cluster system is configured by the computers C1 to C4. In this cluster system, the cluster control unit (CS1) 10 operates. The cluster control unit 10 is realized by a cluster control program (cluster software) (not shown) provided in each of the computers C1 to C4 operating synchronously and integrally while communicating with each other. Virtual machine. For this reason, the cluster control unit 10 can be considered to exist across the computers C1 to C4. The cluster control unit 10 includes an optimal service allocation unit 11 that realizes an optimal service allocation function, a service reallocation unit 12 that realizes a service reallocation function, a policy management unit 13 that realizes a policy management function, and a load that realizes a load management function. It has the management part 14 and the service control part 15 which implement | achieves a service control function.

  When the service needs to be activated, the service optimal arrangement unit 11 determines a computer that is optimal for service execution according to the policy information stored by the policy management unit 13. Specifically, the policy information specifies, for example, the following items (1) to (5) (rules for operation).

(1) Service priority.

  A priority is given to execution for each service. According to the priority of the service, the order in which necessary resources, that is, computers are allocated is determined. In addition, a service with a low priority may be stopped in order to execute a service with a high priority.

(2) Priorities of computers assigned to services.

  When there are a plurality of computers that can execute the service, a computer is assigned a priority.

(3) Relationship between services (exclusive, dependent, etc.).

  A service that cannot be executed at the same time is called a service in an exclusive relationship, and a service that can be executed only when another service is executed is a service in a dependency relationship. Also, a service that cannot be executed on the same computer is called a server exclusive service, and a service that can be executed only when another service is being executed is a server dependent service. Such a relationship between services is set.

(4) Allocation of essential resources (peripheral devices, etc.) for service execution.

  Set the resources required to execute the service, and set the service so that it will not be executed on computers other than those with that resource.

(5) Computer load status (assigned to the computer with the lowest load status).

  A condition such as selecting a computer with the lowest load when executing a service and selecting a computer that does not become overloaded even when the service is executed is set.

  The service relocation unit 12 is an element related to the gist of the present embodiment. When an imbalance occurs in the computer arrangement of the service due to a change in the service load status or a failure that does not cause the computer to stop, the service relocation unit 12 Is determined according to the policy information stored by the policy management unit 13.

  The policy information related to the rearrangement specifies the following items (1) to (4), for example.

(1) Enable / disable switchover of own service.

  Stopping a service being executed and transferring the stopped service to another computer in order to continue execution on another computer is referred to as a switchover. Set whether to enable or disable this switchover. This may be set statically in advance or may be dynamically set so that critical processing is disabled during execution.

(2) When there is no node that can execute the service, other services can be stopped.

  When there is no computer that can execute this at the time of starting a certain service, it sets whether or not it is possible to stop a running service having a lower priority than the service and start the service.

In this case, the stopped service may be set to switch over to another computer. These settings can be set for the entire system, for each service, or for each computer.

(3) Switchover / stop service criteria (high load priority / low load priority).

Examples of criteria include
If you want to switch over / stop first after a high-load service,
If you want to switch over / stop first from a low-load service,
The priority of services to be switched over or stopped may be set. Such a setting can be set for each system and each computer.

In addition, in order to enable / disable the switchover of the last one remaining service, it is necessary to consider the relationship between the size of the service and the capacity of the computer. For example, even if an overloaded service is switched over to a computer having the same capacity for a certain computer, it is still overloaded. In this case, the switchover is impossible.

(4) Response when load conditions change.

  Sets whether to perform service switchover / stop, etc. when the load status of the computer changes. The load status can be set as a change threshold.

  (4-1) Service relocation is executed to the extent that service switchover / stop does not occur in the case where current status is important.

  (4-2) In a case where optimum arrangement is emphasized, the service is rearranged so that it is optimal even if the service is switched over or stopped.

  In addition to this, the service relocation unit, which will be described later, detects the necessity, for example, when a problem occurs in a computer that does not stop, and the capacity temporarily decreases. Is done.

  These policy information can be set in advance by the user. The service determined to be rearranged is stopped by the service optimum arrangement unit 11 until the computer to be executed is assigned.

  The policy management unit 13 stores and manages policy information used by the service optimum arrangement unit 11 and the service rearrangement unit 12.

The load management unit 14 determines the service load and the computer load status in each of the computers C1 to C4,
If it is necessary to relocate the service based on the determination result, the service relocation unit 12 is notified of the fact along with the load information. Upon receiving this notification, the service rearrangement unit 12 executes a service rearrangement process as described later.

  The load information includes the usage amount and response time of the CPUs, memories, and disks of the computers C1 to C4. The computers C1 to C4 have node load monitors 21 to 24, respectively, and monitor their load conditions.

(Cluster system operation)
The cluster system 10 manages the execution of the parallel execution type service and the high availability type service created by the user. The parallel execution type service is, for example, a Web service, and is a type of service that can be simultaneously executed by a plurality of computers C1 to C4 at a time. The number of services to be executed at one time by the parallel execution type service is managed by the load management unit 14, and the number of services increases when the load is high, and the number of services decreases when the load is low.

  On the other hand, the highly available service created by the user is a database search service, for example, and is a service that can be executed by only one computer (for example, C2) at one time. The high availability service is created so as to continue processing by moving to another computer when a failure occurs, or when a failure is predicted or a switchover occurs during a high load.

  For example, if the load management unit 14 of the cluster system 10 determines that the load of the computer C2 is close to the limit when the load of the highly available service being executed on the computer C2 increases rapidly, the service needs to be relocated. Is notified to the service relocation unit 12.

  The service relocation unit 12 starts service relocation processing for a highly available service or a parallel execution type service according to policy information stored in the policy management unit 13 (which can be set by the user).

  Specifically, for example, when the service rearrangement unit 12 determines the rearrangement of the parallel execution type service, the service control unit 15 temporarily stops the parallel execution type service in response to the determination. After the parallel execution type service is stopped, the service optimal arrangement unit 11 selects a computer (for example, C1) optimal for service execution. The service control unit 15 on the selected computer (for example, C1) executes a service automatic switchover by starting a parallel execution type service.

  The service automatic switchover mechanism by the cluster system 10 as described above makes it possible to optimally arrange services corresponding to dynamic load fluctuations.

(Service allocation processing)
The procedure of the service rearrangement process of the cluster system 10 of this embodiment will be described below with reference to the flowchart of FIG.

  The service rearrangement unit 12 makes an inquiry to the policy management unit 13 and executes the rearrangement process according to the setting of policy information set by the user, for example. As described above, the policy information specifies the following items (1) to (4), for example.

(1) Enable / disable switchover for each service.

(2) When there is no node that can execute the service, other services can be stopped.

(3) Criteria for switchover / stop service.

  (3-1) High load priority / low load priority.

  (3-2) Whether or not the last service can be switched over.

(4) Response when load conditions change.

  (4-1) Relocation to the extent that service is not interrupted in the case where current maintenance is important.

  (4-2) In the case where optimum placement is emphasized, rearrangement is performed while causing a service stop.

  As described above, the load management unit 14 determines whether or not it is necessary to rearrange services in accordance with the determination of the load status (step S1). Examples of this criterion include, for example, “when a computer has a continuously high load and service execution delay is predicted” or “a computer has a high load (predicted) high priority service that is waiting to be executed. It is determined that it is necessary to rearrange services.

  In the following, processing when service relocation is necessary (YES in step S1) will be described.

  The service relocation unit 12 determines whether there is a service that can be switched over or stopped according to the policies (1 and 3) of the policy information (step S2). If the determination result is “YES”, the service control unit 15 of the cluster system 10 performs a service switchover until a service having a lower priority than the service set to be switchable is not required to be relocated. (Step S3).

  On the other hand, when there is no service that can be switched over, the service rearrangement unit 12 determines whether or not forcible treatment is possible according to policy (2) of the policy information (NO in step S2, S4). If the forcible treatment is possible, the process proceeds to the process of executing the switchover from the one with the lower priority until the relocation of the service becomes unnecessary (YES in step S4, S3).

  If the forcible treatment cannot be performed, the cluster system 10 searches for an available provisioning computer (standby computer), and if it exists, adds the computer C5 (NO in steps S4, S5, and S6). Here, if the provisioning computer C5 is designated to be returned when the load on the computer system is reduced, the provisioning computer C5 is returned when the load is reduced. If there is no provisioning computer that can be used, a return is made after a certain period of sleep (NO in step S5, S11).

  Next, a case where the rearrangement of services is unnecessary according to the determination result of the load management unit 14 will be described (NO in step S1).

  In accordance with the policy information policy (4-2), the service rearrangement unit 12 executes the service rearrangement process when the load on the optimized arrangement is emphasized (YES in step S7, YES in S8). . Otherwise, the service rearrangement process ends (NO in step S7, NO in S8).

  Here, whether or not the computer is becoming heavily loaded is determined by whether or not the load averaged at regular intervals is increasing monotonously and it can be predicted that the load will be high in the not-too-distant future. it can.

  Further, when executing the service rearrangement process, the service rearrangement unit 12 determines whether or not the movement of the service is more optimal, and executes the switchover of the service if optimal (step S9). YES, S10). If it cannot be determined that the arrangement is optimal, the service rearrangement process ends (NO in step S9).

  Here, the criterion for determining the optimal arrangement is when the service to be relocated on the selected computer is operated with the same load as the current load, and the load state between the computers is more averaged. In addition, there is a case where it is considered that it is faster to perform processing with the selected computer even if the overhead of service switchover is taken into account.

  Here, as a policy for service relocation, a switchover policy can be enabled / disabled for each service, a policy that emphasizes the maintenance of the current status can be performed, and even if it is stopped by a switchover, it is executed as long as it can not be started by the switchover destination computer By not doing so, it becomes possible to prevent the switchover from being repeated in response to the load fluctuation of the computer.

  In short, with the cluster system of this embodiment, by providing a service relocation function managed on a policy basis, it is possible to relocate services according to dynamic load status changes, and users It is possible to easily construct a cluster system suitable for the operational environment.

(Second Embodiment)
FIG. 3 to FIG. 5 are block diagrams showing the system configuration of the computer system related to the second embodiment and its changes.

  As shown in FIG. 3, the computer system in the initial state has a configuration in which, for example, five computers C1 to C5 are connected to each other on the network N. Further, on the network N, a sixth computer C6 is connected. The computer C6 is stopped and registered in the provisioning computer pool 60 as a provisioning computer (standby computer).

  The provisioning computer pool 60 conceptually illustrates and generically indicates that one or more stopped computers are registered as provisioning computers.

  Registering a computer as a provisioning computer in the provisioning computer pool 60 means registering information related to a provisioning computer (not shown) such as a processor name and a MAC address as registration information. Manage multiple registered provisioning computers.

  The computers C1 to C3 are operating under the control of the operating system OS (OS-1-1 to OS-1-3), respectively. The computers C4 and C5 are operating under the control of the operating system OS (OS-2-1, OS-2-2).

  In the operating computers C1 to C5, a provisioning computer allocation unit 31 that realizes a provisioning computer allocation function, a provisioning computer disconnection unit 32 that realizes a provisioning computer disconnection function, and a provisioning policy management unit (hereinafter referred to as provisioning policy management function). 33, which may simply be abbreviated as policy management unit). The computer C1, the computer C2, and the computer C3 are operated by the provisioning computer allocation unit 31, the provisioning computer disconnection unit 32, and the provisioning policy management unit 33, respectively. Computer C2 and Computer C3 constitute a cluster system CS1. Reference numeral 30 schematically illustrates the cluster system CS1. On the other hand, the computer C4 and the computer C5 are operated by the provisioning computer allocation unit 31, the provisioning computer disconnection unit 32, and the provisioning policy management unit 33, respectively. The computer C5 constitutes the cluster system CS2. Reference numeral 40 schematically illustrates the cluster system CS2. Each of these cluster systems is unrelated to each other and does not relate services to each other.

  A plurality of storage devices (disk devices) 50 to 57 and 70 are connected to the computer system via a storage area network SAN (Storage Area Network) 45.

  In this computer system, a boot image for starting each computer is stored and registered in advance in storage devices (disk devices) 50-57. Here, the boot image includes an operating system for starting the computer and an application program executable by the operating system.

  The storage devices 50 to 53 and 54 to 57 have boot images OS-1-1, OS-1-2, OS-1-3, OS-1-4, OS-2-1, OS-2-2, respectively. 2, OS-2-3 and OS-2-4 are registered. For example, a boot image (OS-1-3) for starting up the computer C3 is registered on the storage device 52. When the computer C3 is activated using this boot image (OS-1-3), the computer C3 becomes an active computer whose operation is controlled by the OS (OS-1-3). In FIG. 3, which computer is booted with which boot image is indicated by an arrow.

  On the other hand, as shown in FIG. 5, a boot image (OS-2-4) for starting up the computer C3 is registered on the storage device 57. When the computer C3 is started using this boot image (OS-2-4), the computer C3 becomes an active computer whose operation is controlled by the OS (OS-2-4). In FIG. 5, which computer is booted with which boot image is indicated by an arrow.

(Cluster system operation)
The provisioning computer allocating unit 31 provisions stored in a provisioning policy database (hereinafter abbreviated as policy DB) 70 that can be accessed via the policy management unit 33 when computers to be executed by the cluster control units 30 and 40 become necessary. A provisioning computer is allocated to the cluster system according to the policy information.

  The provisioning computer disconnection unit 32 disconnects the computers in the cluster system according to the policy DB 70 that can be accessed via the policy management unit 33 when a surplus occurs in the computers executed by the cluster control units 30 and 40, and pools the provisioning computer as a provisioning computer. 60.

  The policy management unit 33 provides a setting / reference function for provisioning policy information (hereinafter sometimes simply referred to as policy information). The policy information specifies, for example, the following provisioning policies (1) to (4).

(1) Computer allocation level (priority) for each cluster system.

  When provisioning computer requests are received from two or more cluster systems at the same time, the order (priority) of the cluster systems to be preferentially assigned is set. When there is no provisioning node required when a request is made from the cluster system, a computer assigned to the cluster system having a low priority may be forcibly assigned to the requested cluster system.

(2) Whether the provided computer can be returned or not.

Whether or not the provisioning computer assigned in the cluster system can be converted into a provisioning pool is set. Therefore, if this setting is not possible, the number of assigned computers in the cluster system will only increase.

(3) Whether the provided computer can be forcibly returned.

  Sets whether the computer provided to the cluster system from the provisioning pool can be forcibly returned. In other words, whether or not the system is forcibly returned does not interfere with system operation. For example, when there is a spare computer in the provisioning pool when there is a request from a cluster system with a high priority, a forced return request is made to the cluster system with a low priority.

(4) Index of the number of provided computers (required number of computers, maximum number of computers, number of initial computers).

  The number of computers necessary for configuring the cluster system is set as the required number of computers. The maximum number of computers that can be allocated to the cluster system is defined as the maximum number of computers. Also, the optimal number of computers assigned at the time of starting the cluster system is set as the initial number of computers. In this way, it is possible to set an index for determining the number of computers to be provided to the cluster system.

  The policy information is normally set in the policy DB 70 when the user constructs / maintains the computer system.

  FIG. 8 shows an example of provisioning policy information registered in the provisioning DB 70 for registration in each computer in the cluster system shown in FIG.

(Provisioning computer allocation process)
The procedure of the provisioning computer assignment process of this embodiment will be described below with reference to the flowchart of FIG.

  First, as shown in FIG. 3, in the computer system in the initial state, the computers C1 to C3 are operating, and the cluster system (CS1) 30 is operating. The computers C4 and C5 are in operation, and the cluster system (CS2) 40 is in operation. Furthermore, the computer C6 is stopped and registered in the pool 60 as a provisioning computer.

  Here, when the load on the cluster system (CS2) 40 increases and the two computers C4 and C5 cannot process, the cluster system (CS2) 40 requests the provisioning computer allocation unit 41 to add a computer (step S21). YES)

The provisioning computer allocation unit 41 searches the provisioning computer pool 60, and
The registered computer C6 is taken out and added to the requested cluster system (CS2) 40 (YES in step S23, S24). Here, as shown in FIG. 4, the provisioning computer allocating unit 41 sends an unused boot image (OS-2-3) from the storage device 56 among the boot images belonging to the cluster system (CS2) 40. Take out, connect to the computer C6 and start.

  However, if the requirements to be satisfied by the boot image are specified in detail from the cluster system (CS2) 40, a boot image that meets the requirements is retrieved.

  By the way, when a computer addition request is made simultaneously from the two cluster systems 30 and 40, the provisioning computer allocation unit 31 and 41 accesses the policy DB 70 via the policy management units 33 and 43, and follows the policy information. A cluster control unit with a high computer allocation level is selected (step S22). For example, when the allocation level is higher in the cluster system (CS2) 40, the provisioning computer assignment unit 41 searches the provisioning computer pool 60 and preferentially assigns the registered computer C6 (in step S23). YES, S24).

  Further, when the load on the cluster system (CS2) 40 further increases and the processing cannot be performed by the three computers C4 to C6, the cluster control unit 40 requests the provisioning computer allocation unit 41 to add a computer.

  Since the computer is not registered in the provisioning computer pool 60, the provisioning computer allocation unit 41 determines whether there is a cluster control unit that can be forcibly returned according to the policy information (NO in step S23, S25). . If it does not exist, it goes through a sleep state for a certain time and enters a standby state until the computer is registered in the pool 60 (NO in step S25, S26).

  On the other hand, for example, when the cluster system (CS1) 30 can be forcibly returned, the provisioning computer allocation unit 41 requests the computer on the cluster system (CS1) 30 to forcibly return (YES in step S25). The provisioning computer disconnection unit 32 on the computer of the cluster system (CS1) 30 requested for forced return determines a computer (for example, C3) that can be separated and registers it as a provisioning computer in the provisioning computer pool 60 (step S27).

  When the computer C3 disconnected from the cluster system (CS1) 30 is registered in the provisioning computer pool 60, the provisioning computer allocation unit 41 of the cluster system (CS2) 40 searches and registers the provisioning computer pool 60. The computer C3 is taken out and assigned (YES in step S23, S24).

  As shown in FIG. 5, the provisioning computer allocating unit 41 extracts a boot image (OS-2-4) that is not used from the storage device 57 among the boot images belonging to the cluster system (CS2) 40, and calculates the computer. Connect to C3 and start.

(Provisioning computer disconnection process)
Next, the procedure of the provisioning computer disconnection process according to this embodiment will be described with reference to the flowchart of FIG.

  Here, upon receiving the computer disconnection request, the provisioning computer disconnection unit 32 of the cluster system (CS1) 30 determines a computer (here, C3) that can be disconnected on the cluster system (CS1) 30 according to the policy information (step S3). S31 YES, S33).

  Further, the provisioning computer disconnection unit 32 issues a switchover request to the service running on the determined computer C3 (step S34). Here, in the cluster control unit 30, when all services are waiting to be stopped according to the policy information, the provisioning computer disconnection unit 32 waits for all services to stop, disconnects the computer C 3, and becomes a provisioning computer. Registration is made in the provisioning computer pool 60 (YES in steps S35, S37, S38).

  On the other hand, if the disconnection condition is not waiting for the stop of all services, the provisioning computer disconnection unit 32 waits for a certain time as preparation for disconnection, disconnects the computer C3, and registers it as a provisioning computer in the provisioning computer pool 60 (step NO of S35, S36, S38).

  As described above, according to the present embodiment, when there is a provisioning computer addition request from a plurality of cluster systems, for example, from the cluster system (CS1) 30 in which forced return is set, according to the policy information, It is possible to execute a process of allocating a computer to a cluster system (CS2) 40 having a high computer allocation level. In short, by having a provisioning computer assignment / separation function capable of setting a provisioning policy for each cluster system, it is possible to assign (move) an optimum computer based on the computer assignment level between the cluster systems. By linking such a cluster system with, for example, a billing system, it is possible to construct a system that realizes an advanced service level agreement (SLA) in a network service.

  Various aspects of this embodiment can be summarized as follows.

(1) In a computer system in which two or more computers are connected to realize two or more cluster systems,
At least one provisioning computer that can be commonly used by the respective cluster systems;
Policy management means for storing policy information for designating a policy for assigning or detaching a provisioning computer in a changeable manner;
A computer system comprising: an allocation process for allocating an additional request computer from the at least one provisioning computer in accordance with the policy information, or an allocating / detaching unit for detaching a surplus computer.

  (2) The allocation / detachment means allocates a computer registered as at least one provisioning computer according to the policy information, or a computer used in another cluster system to a necessary cluster system in the item (1). The computer system described.

  (3) The computer system according to (1), wherein the allocation / detachment unit disconnects a computer used in a cluster system according to the policy information and registers it as the at least one provisioning computer.

  (4) The policy management means manages a database that stores the policy information in a changeable manner, and extracts or sets the policy information from the database in accordance with access from the computers. The computer system described in 1.

(5) A program that is applied to a computer system to which two or more computers are connected, and is included in each of two or more cluster systems,
A procedure of executing a process of assigning a computer of an additional request from at least one provisioning computer that can be commonly used by each cluster system in accordance with the changeable policy information;
A program for causing the computer system to execute a procedure for executing a process of disconnecting at least one provisioning computer used by each cluster system according to the policy information.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the components without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

1 is a block diagram showing a system configuration relating to a first embodiment of the present invention. The flowchart for demonstrating the procedure of the service rearrangement process regarding 1st Embodiment. The block diagram which shows the system configuration | structure regarding 2nd Embodiment. The block diagram which shows the change of the system configuration regarding 2nd Embodiment. The block diagram which shows the change of the system configuration regarding 2nd Embodiment. 10 is a flowchart for explaining a procedure of a provisioning computer assignment process according to the second embodiment; The flowchart for demonstrating the procedure of the separation process of the provisioning computer regarding 2nd Embodiment. The figure which shows an example of the provisioning policy information regarding 2nd Embodiment.

Explanation of symbols

10 ... cluster control unit, 11 ... service optimal arrangement unit,
12 ... Service relocation unit, 13 ... Policy management unit, 14 ... Load management unit,
15 ... Service control unit, 30 ... Cluster system CS1,
31 ... provisioning computer allocation unit,
32 ... provisioning computer disconnection unit, 33 ... provisioning policy management unit,
40 ... Cluster system CS2,
50 to 57: Storage device (boot image registration disk),
60 ... Provisioning computer pool,
70... Provisioning policy database (policy DB), C1 to C6.

Claims (9)

In a computer system with two or more computers,
The priority of the service for determining the allocation of a plurality of services executed by the computer, the priority of the computer allocated to the service, the exclusion between the services, the relationship including the dependency relationship, and the resource including the peripheral device essential for the service execution Policy information for optimal placement having at least one of the allocation and the load status of the computer for allocating the service to the computer with the lowest load status, service switchover enabled / disabled, and no service executable node Policy management unit for storing service relocation policy information having at least one of whether or not other services can be stopped and whether or not the switchover / other services stop determination criteria and the response when the load status changes When,
A service optimal allocation unit that allocates a service to be executed based on the policy information for optimal allocation to an optimal computer;
A load management unit that determines whether or not the relocation of the service is necessary depending on the service load and the computer load state in each computer after the optimal placement;
A service for determining an optimal computer for execution of a service determined to be relocated by the load management unit based on the relocation policy information according to a determination result that the service needs to be relocated by the load management unit A relocation section;
The execution of the service determined to be necessary for the rearrangement is temporarily stopped, and the rearrangement is required by a computer optimal for the execution of the service determined to be necessary for the rearrangement determined by the service allocation unit. And a service control unit that activates the determined service and switches over the service. The said service optimal arrangement | positioning part determines the optimal computer for the execution at the time of starting of a desired service with reference to the policy information for optimal arrangement | positioning preserve | saved in the said policy management part. Computer system. The service relocation unit includes means for detecting the necessity of service relocation when an imbalance occurs in the service arrangement being executed between computers, and refers to the policy information for relocation according to the output of the detecting means and wherein the rearranging of the service, computer system according to claim 1. The computer system according to claim 3, wherein the detecting unit is included in the load management unit as a detecting unit that detects a load state of each computer. 5. The computer system according to claim 4, wherein the means for detecting includes a node load monitor of each computer. The service relocation unit determines the necessity of service relocation according to a change in load status of each computer,
2. The computer system according to claim 1, wherein when there is a need for relocation of the service, relocation processing including use of a spare computer is executed in accordance with the relocation policy information. Further, the computer system includes two or more cluster systems and a provisioning computer group that can be commonly used by each cluster system,
  The policy management unit stores policy information for allocation and detachment processing for designating a policy for allocation processing and detachment processing of the provisioning computer,
  2. The computer system further includes an allocation / separation unit that executes an allocation process for allocating an additional request computer from the provisioning computer group or a detachment process for detaching a surplus computer in accordance with the policy information for allocation and detachment processing. The computer system described in 1. A service execution method by a computer system in which two or more computers are connected to realize one cluster system,
The priority of the service for determining the allocation of a plurality of services executed by the computer, the priority of the computer allocated to the service, the exclusion between the services, the relationship including the dependency relationship, and the resource including the peripheral device essential for the service execution Policy information for optimal placement having at least one of the allocation and the load status of the computer for allocating the service to the computer with the lowest load status, service switchover enabled / disabled, and no service executable node A service relocation policy information having at least one of whether or not the other service can be stopped, the switchover / other service stop determination criterion, and the response when the load status changes ,
A process of allocating a service to be executed based on the optimal placement policy information to an optimal computer ;
Determine whether or not it is necessary to relocate the service according to the service load and the computer load status on each computer after the optimal placement,
In accordance with the determination result that the load management unit needs to relocate the service, based on the relocation policy information, determine a computer optimal for executing the service determined to be relocated by the load management unit,
The execution of the service determined to be necessary for the rearrangement is temporarily stopped, and the rearrangement is required by a computer optimal for the execution of the service determined to be necessary for the rearrangement determined by the service allocation unit. A service execution method characterized by starting a determined service and switching over the service. A program for realizing a cluster system by applying it to a computer system in which two or more computers are connected,
The priority of the service for determining the allocation of a plurality of services executed by the computer, the priority of the computer allocated to the service, the exclusion between the services, the relationship including the dependency relationship, and the resource including the peripheral device essential for the service execution Policy information for optimal placement having at least one of the allocation and the load status of the computer for allocating the service to the computer with the lowest load status, service switchover enabled / disabled, and no service executable node A procedure for storing policy information for relocation of a service having at least one of whether or not the other service can be stopped, whether the switchover / other service is stopped, and a response when the load status changes ,
A procedure for executing a process of assigning a service to be executed based on the policy information for optimal placement to an optimal computer ;
A procedure for determining whether or not relocation of services is necessary depending on the service load and the computer load status in each computer after the optimal placement;
A procedure for determining an optimal computer for executing a service determined to be relocated by the load management unit based on the relocation policy information in accordance with a determination result that the load management unit needs to relocate the service. When,
The execution of the service determined to be necessary for the rearrangement is temporarily stopped, and the rearrangement is required by a computer optimal for the execution of the service determined to be necessary for the rearrangement determined by the service allocation unit. Start the determined service and switch over the service;
For causing the computer system to execute .

Images