JP5471292B2 - Virtual machine movement control program, virtual machine movement control device, and virtual machine movement control method - Google Patents

Description

  The present invention relates to a virtual machine movement control program, a virtual machine movement control device, and a virtual machine movement control method for controlling movement of a virtual machine that moves a virtual machine on a certain physical machine to another physical machine.

  In a virtual machine system in which a plurality of virtual machines operate on a plurality of physical machines, there is a technique for moving a virtual machine that moves a virtual machine on a certain physical machine to another physical machine. The virtual machine movement is performed, for example, by copying the image of the virtual machine on the memory of the movement source physical machine to the memory of the movement destination physical machine.

  The virtual machine movement is performed for the purpose of, for example, recovery from a physical failure, efficient resource allocation, and separation of a high-load virtual machine that adversely affects other virtual machines in a virtual machine system. In a virtual machine system in which virtual machine movement is performed, the load on each virtual machine is monitored, and the virtual machine is moved to another physical machine before the total resources of the physical machine are saturated. Is avoided.

  As a technique for moving a virtual machine, a technique called live migration that moves between physical machines while a virtual machine is operating has been developed and used. With live migration technology, you can move virtual machines without stopping the services of the virtual machines to be moved.

  A technique for changing the assignment of blade servers in accordance with the service load is known. Also, a technology for adjusting resources by moving a virtual machine or the like is known. In addition, a technique for effectively utilizing server resources by scaling the grid configuration according to the load is known.

JP 2004-110791 A Special table 2007-508623 Special table 2008-527514

  In the technology of live migration, it is known that a virtual machine system consumes a non-negligible amount of resources when executing virtual machine migration (see Reference 1).

[Reference 1]
W. Huang, Q. Gao, J. Liu, and DK Panda (Ohio State University), Jiuxing Liu (IBM TJ Watson Research Center), “High Performance Virtual Machine Migration with RDMA over Modern Interconnects”, IEEE International Conference on Cluster Computing (Cluster'07), Austin, TX, September 2007, Selected as a BEST Paper., Http://nowlab.cse.ohio-state.edu/publications/conf-papers/2007/huang-cluster07.pdf
Therefore, if live migration is performed when physical machine resources are saturated, the CPU (Central Processing Unit) resources necessary for live migration are insufficient, which increases the time required to move the virtual machine. In addition, the service level of a running virtual machine may be degraded.

  As an opportunity to start live migration, for example, the following may be considered.

  [A] When the margin of physical machine resources becomes zero, live migration is started (zero threshold).

  [B] When the physical machine resource margin is reduced to the cost required for live migration, live migration is started (live migration load threshold).

  However, in the case of [A], the virtual machine movement starts when the physical machine resources are saturated, and the original service performance is reduced by the live migration load.

  In the case of [B], resources for the live migration load are always reserved, resulting in poor resource utilization efficiency. In addition, even when the live migration load threshold is slightly exceeded, even when there is no need for virtual machine movement, unnecessary virtual machine movement occurs.

  The present invention solves the above-described problems and makes it possible to secure resources necessary for moving a virtual machine at the time of moving the virtual machine without reserving resources necessary for moving the virtual machine in advance. The purpose is to provide technology.

  The disclosed program causes a computer that controls movement of a virtual machine between physical machines to function as follows in a virtual machine system having a plurality of physical machines on which the virtual machine can operate.

That is, the program includes a resource allocation information including information on a resource allocation amount for a virtual machine for each physical machine, stored in a storage device accessible by the computer, on a computer on which the program is installed and executed, based on the grouping information machine for each virtual machine that performs the same process with the load balancing, resource allocation virtual machines belonging to other groups that run on the mobile object moves virtual machine that runs Domoto physical machine and the amount of reduction that is the object subtractive adjusted virtual machines, increasing adjusted virtual machine is an increase target resource quotas virtual machines running reduced adjusting virtual machine and belonging to the same group and the migration source physical machine other than the physical machine and to a procedure for each selection, a predetermined required to move the moving target virtual machine to increase adjusted virtual machine Assignment by adding a source volume, have rows adjustment to reduce the predetermined amount of resources from the reduced adjustment virtual machine, ensure a predetermined resource amount as a resource for moving the Oite movement target virtual machine to the migration source physical machine after the adjustment And execute the procedure.

  With the above technology, resources necessary for moving a virtual machine can be secured as required when the virtual machine is moved. For this reason, even when the physical machine is saturated, the virtual machine is moved without affecting the performance of the entire service by the virtual machine. In addition, since it is not necessary to always secure resources necessary for moving the virtual machine, the resources of each physical machine can be effectively used.

It is a figure which shows the structural example of the data center system by this Embodiment. It is a figure (1) explaining the movement of the instance by this Embodiment. It is a figure (2) explaining the movement of the instance by this Embodiment. It is a figure which shows the structural example of the migration control part by this Embodiment. It is a figure which shows the example of the resource management data by this Embodiment. It is a figure which shows the example of the migration resource allocation range data by this Embodiment. It is a figure which shows the structural example of the temporary resource allocation data by this Embodiment. It is a migration resource ensuring process flowchart by the migration control part of this Embodiment. It is a movement object affiliation cluster instance selection process flowchart by the movement object affiliation cluster instance selection part of this Embodiment. It is a non-move target cluster instance selection process flowchart by the non-move cluster instance selection unit of the present embodiment. It is a migration resource securing determination processing flowchart by the migration resource securing determining unit, the migration target belonging cluster instance selecting unit, and the non-moving target belonging cluster instance selecting unit according to the present embodiment. It is a migration resource release process flowchart by the migration resource release part of this Embodiment. In Example 1, it is a figure which shows the state before execution of migration. In Example 1, it is a figure which shows the state by which the migration resource was ensured. In Example 1, it is a figure which shows the state of migration execution. FIG. 10 is a diagram illustrating a state where migration resources are released in the first embodiment. In Example 2, it is a figure which shows the state before execution of migration. In Example 2, it is a figure which shows the state by which the migration resource was ensured. In Example 2, it is a figure which shows the state of migration execution. In Example 2, it is a figure which shows the state by which the migration resource was released.

  Hereinafter, the present embodiment will be described with reference to the drawings.

  FIG. 1 is a diagram illustrating a configuration example of a data center system according to the present embodiment.

  A data center system 1 shown in FIG. 1 is a data center system that provides a hosting service for lending a computer to a tenant. A tenant is a service provider that provides services to end users using a data center hosting service. An end user is a user of a service operated by a tenant.

  The data center of this embodiment uses a virtualization mechanism in order to share a physical machine among a plurality of tenants. That is, the data center system 1 shown in FIG. 1 is a virtual machine system, and the computer that the data center lends to a tenant is a virtual machine. In addition to virtual machines, the data center also provides infrastructure necessary for services such as load balancers and network / internet connections between virtual machines.

  The data center system 1 includes a management mechanism device 10, a physical server 20, an SLB (Server Load Balancer) 30, a service network 40, and a management network 50.

  The management mechanism device 10 is a computer that manages and controls the entire data center system 1 via the management network 50 and the like. The manager of the data center can manage the data center system 1 by operating the management mechanism device 10.

  The physical server 20 is a physical machine that can operate a plurality of virtual machines. The data center system 1 includes a plurality of physical servers 20. The physical server 20 includes a virtualization mechanism control unit 21 and an instance 22.

  The virtualization mechanism control unit 21 is also called a hypervisor or a virtual machine monitor, and manages and controls the instance 22 that operates on the physical server 20. For example, the virtualization mechanism control unit 21 performs resource allocation to the instance 22 operating on the own physical server 20, movement control of the instance 22 between the own physical server 20 and the other physical server 20, and the like.

  The instance 22 is a computer that constitutes a tenant service. That is, in the data center of the present embodiment that lends a virtual machine to a tenant, the instance 22 is a virtual machine that operates on the physical server 20.

  The SLB 30 is a load balancer that distributes access to a tenant service by an end user to a plurality of instances 22 that perform processing of the same service. In the present embodiment, a group of instances 22 that perform the same processing and are connected under the same SLB 30 is referred to as a cluster. That is, the SLB 30 performs load distribution to the instances 22 belonging to the same cluster.

  The tenant terminal 2 is a computer used by a tenant who performs services using the data center system 1 to perform service management. The tenant terminal 2 can access an instance 22 that performs tenant service processing via a service management network 4 such as a VPN (Virtual Private Network). The tenant accesses the instance 22 borrowed from the data center using the tenant terminal 2 and manages the service provided by the tenant.

  The end user terminal 3 is a computer for an end user who wants to receive the tenant service to access the instance 22 that performs the processing of the tenant service. The end user terminal 3 can access the instance 22 that performs processing of the tenant service via the Internet 5 and the service network 40. The end user uses the end user terminal 3 to access the instance 22 that performs processing of the tenant service, and receives a service provided by the tenant using the data center system 1.

  For example, in FIG. 1, it is assumed that a certain tenant borrows an instance 22A of the physical server 20A and an instance 22C of the physical server 20B from the data center to provide services. The tenant accesses the instance 22A and the instance 22C using the tenant terminal 2 and manages services by the instance 22A and the instance 22C, as indicated by the dotted arrows in FIG. In this case, the instance 22A and the instance 22C are instances 22 that perform processing of one and the same service, and thus are instances 22 that belong to the same cluster, as indicated by a broken line frame in FIG. That is, the access from the end user terminal 3 to the service by these instances 22 is distributed to the instances 22A and 22C by the SLB 30, as indicated by solid arrows in FIG.

  In the data center system 1, an instance 22 used by each of a plurality of tenants is executed on one physical server 20. In such an environment, an increase in the load of the instance 22 used by a certain tenant may adversely affect the instances 22 of other tenants sharing the same physical server 20. In the data center system 1, by monitoring the load of each instance 22 and moving the instance 22 to another physical server 20 before the total resources of the physical server 20 are saturated, it is possible to avoid an overload state. ing.

  For example, in FIG. 1, it is assumed that the resources of the physical server 20A are likely to be saturated. At this time, in the data center system 1, the instance 22 is moved from the physical server 20A where resources are likely to be saturated to the physical server 20B where resources are still sufficient. For example, the instance 22B operating on the physical server 20A is moved to the physical server 20B as indicated by the one-dot chain line in FIG.

  As a technique for moving the instance 22 between the physical servers 20, that is, moving the virtual machine, there is a technique called live migration that moves between the physical servers 20 while the instance 22 is operating as described above. In the present embodiment, the instance 22 is moved by the live migration technique without stopping the service by the instance 22 to be moved.

  In live migration, the physical server 20 needs resources for moving the instance 22. Therefore, as described above, when live migration is performed after the physical server 20 is saturated, the resources of the running instance 22 are reduced in order to allocate resources for live migration. There is a problem of end. In addition, as described above, if the physical server 20 reserves resources necessary for live migration in advance, there is a problem that resource utilization efficiency deteriorates.

  In order to deal with such a problem, the data center system 1 according to the present embodiment does not reserve the resources necessary for moving the instance 22 and only needs to move the instance 22 when the instance 22 is moved. Resources are secured.

  The management mechanism device 10 includes a migration control unit 100. The migration control unit 100 manages and controls the movement of the instance 22 for the entire data center system 1. The migration control unit 100 is realized by hardware such as a CPU and a memory provided in the computer of the management mechanism device 10 and a software program.

  When the migration of the instance 22 occurs, the migration control unit 100 performs control for securing on the physical server 20 enough resources for the migration of the instance 22. Hereinafter, a technique for securing resources necessary for moving the instance 22 by the migration control unit 100 according to the present embodiment will be described.

  In the following, the movement of the virtual machine according to the present embodiment, that is, the movement of the instance 22 according to the present embodiment is simply referred to as migration. The resources of the physical server 20 include various resources such as CPU resources and memory resources. In the present embodiment, the resources are mainly referred to as CPU resources.

  FIG. 2 is a diagram (1) for explaining the movement of an instance according to the present embodiment.

  In FIG. 2, it is assumed that the thick and thin long frame in each physical server 20 represents the resource amount that each physical server 20 has. Further, it is assumed that the thin line frame on the thick line frame representing the resource amount of each physical server 20 represents each instance 22 operating on each physical server 20 and its resource allocation amount. A cluster to which each instance 22 belongs is shown in parentheses in a thin line frame representing each instance.

  FIG. 2A shows a state before execution of migration.

  In FIG. 2A, it is assumed that the instance 22a is the migration target instance 22 in the current migration. As shown in FIG. 2A, the instance 22a belongs to the cluster a.

  As shown in FIG. 2A, the physical server 20 on which the migration target instance 22a operates becomes the migration source physical server 20b in the current migration. The migration destination physical server 20a is the physical server 20 that is the migration destination of the migration target instance 22a.

  As shown in FIG. 2A, the migration source physical server 20b is in a state where resources are saturated by the instance 22a and the instance 22b, and resources necessary for migration remain in the migration source physical server 20b. Not. If the resource allocation amount of the instance 22a and the instance 22b is reduced in order to generate resources necessary for the migration, the quality of service by the instance 22a and the instance 22b is deteriorated.

  The migration control unit 100 performs processing for securing resources necessary for migration on the migration source physical server 20b without causing degradation in service quality due to migration of the instance 22. Specifically, in the example of the present embodiment shown in FIG. 2, the migration control unit 100 is a physical server 20 on which the instance 22 belonging to the same cluster a as the instance 22a operates, and the physical server 20 having sufficient resources. Perform a search. Here, it is assumed that the migration control unit 100 detects the physical server 20 on which the instance 22c belonging to the same cluster a as the instance 22a operates as the physical server 20c with a margin.

  In the present embodiment, it is assumed that the instance 22a and the instance 22c belonging to the same cluster a are instances 22 that provide the same tenant service. As shown in FIG. 2A, the SLB 30 performs load distribution so that the load of the instance 22a and the load of the instance 22c are the same during normal operation in the load distribution of the cluster a.

  FIG. 2B shows securing resources necessary for migration.

  In the present embodiment, the resource allocation amount of the instance 22 is adjusted in order to secure resources necessary for migration without degrading the quality of service by the instance 22. In the example shown in FIG. 2, migration to the migration source physical server 20b is performed by adjusting the resource allocation amount of the migration target instance 22a on the migration source physical server 20b and the resource allocation amount of the instance 22c on the physical server 20c with a margin. Necessary resources are secured.

  The migration control unit 100 allows the virtualization mechanism control unit 21 of the physical server 20c with sufficient capacity to allocate the resource allocation amount of the instance 22c belonging to the same cluster a as the migration target instance 22a to the resource amount Δ necessary for migration. Instruct to increase only. As shown in FIG. 2B, the virtualization mechanism control unit 21 of the physical server 20c having a margin increases the resource allocation amount of the instance 22c by the amount of resource Δ necessary for migration.

  The migration control unit 100 instructs the SLB 30 to change the setting of the load distribution ratio between the instance 22a and the instance 22c. As shown in FIG. 2B, the SLB 30 changes the setting of the load distribution ratio between the instance 22a and the instance 22c. The changed load distribution ratio is the resource allocation amount of the original instance 22a−Δ: the resource allocation amount of the original instance 22c + Δ.

  The migration control unit 100 reduces the resource allocation amount of the migration target instance 22a by the amount of the resource amount Δ necessary for migration with respect to the virtualization mechanism control unit 21 of the migration source physical server 20b in which the resources are saturated. To instruct. As shown in FIG. 2B, the virtualization mechanism control unit 21 of the migration source physical server 20b reduces the resource allocation amount of the migration target instance 22a by the resource amount Δ required for migration.

  By such control of the migration control unit 100, resources necessary for migration are secured in the migration source physical server 20b. At this time, the resource reduction of the instance 22a for securing resources in the source physical server 20b is dealt with by the increase in resources of the instance 22c that performs the same service processing as the instance 22a. It does not decline.

  FIG. 2C shows execution of migration.

  The virtualization mechanism control unit 21 of the migration source physical server 20b and the virtualization mechanism control unit 21 of the migration destination physical server 20a execute migration according to an instruction from the migration control unit 100. At this time, in the migration source physical server 20b and the migration destination physical server 20a, as shown in the hatched portion of FIG. 2C, resources used for executing migration are allocated. Hereinafter, a resource used for executing migration is referred to as a migration resource.

  Due to the migration, as shown in FIG. 2C, the instance 22a on the migration source physical server 20b is migrated to the migration destination physical server 20a.

  FIG. 2D shows restoration of resource allocation after migration.

  At the stage where the migration shown in FIG. 2C is completed, the resource allocation amount of the instance 22a on the migration destination physical server 20a remains the original resource allocation amount -Δ, and the instance 22c on the physical server 20c with margin is available. The original resource allocation amount + Δ remains the same. Although it is possible to operate the instance 22a and the instance 22c as they are, in this embodiment, the migration control unit 100 restores the resource allocation amounts of the instance 22a and the instance 22c whose resource allocation amounts have been changed.

  The migration control unit 100 causes the virtualization mechanism control unit 21 of the destination physical server 20a to restore the resource allocation amount of the instance 22a to the original resource allocation amount, that is, to reduce the resource allocation amount of the instance 22a by Δ minutes. Instruct to increase only. As shown in FIG. 2D, the virtualization mechanism control unit 21 of the migration destination physical server 20a increases the resource allocation amount of the instance 22a by Δ.

  The migration control unit 100 instructs the SLB 30 to restore the setting of the load distribution ratio between the instance 22a and the instance 22c. As shown in FIG. 2D, the SLB 30 restores the setting of the load distribution ratio between the instance 22a and the instance 22c. The restored load distribution ratio is the resource allocation amount of the original instance 22a: the resource allocation amount of the original instance 22c.

  The migration control unit 100 allows the virtualization mechanism control unit 21 of the physical server 20c with sufficient capacity to restore the resource allocation amount of the instance 22c to the original resource allocation amount, that is, the resource allocation amount of the instance 22c is Δ minutes. Instruct to reduce only. As shown in FIG. 2D, the virtualization mechanism control unit 21 of the physical server 20c having a margin reduces the resource allocation amount of the instance 22c by Δ.

  By such control of the migration control unit 100, the resource allocation amount of the instance 22a and the instance 22c whose resource allocation amount has been changed to secure the migration resource is restored to the state before the occurrence of migration.

  FIG. 3 is a diagram (2) for explaining the movement of the instance according to the present embodiment.

  In the example shown in FIG. 2, the migration resource is secured in the source physical server 20b by adjusting the resource allocation amount of the instance 22 within the cluster a to which the instance 22a to be moved belongs. However, there may be a case where there is no instance 22 belonging to the same cluster as the instance 22a to be moved. In addition, even if there is an instance 22 that belongs to the same cluster as the migration target instance 22a, the physical server 20 on which the instance 22 that belongs to the same cluster as the migration target instance 22a operates may not have sufficient resources.

  Here, an example in which migration resources are secured in the migration source physical server 20 by adjusting the resource allocation amount of the instances 22 belonging to clusters other than the cluster to which the migration target instance 22 belongs will be described with reference to FIG. To do.

  In FIG. 3, it is assumed that the thick and thin long frame in each physical server 20 represents the resource amount of each physical server 20 as in the example shown in FIG. 2. Further, it is assumed that the thin line frame on the thick line frame representing the resource amount of each physical server 20 represents each instance 22 operating on each physical server 20 and its resource allocation amount. A cluster to which each instance 22 belongs is shown in parentheses in a thin line frame representing each instance.

  FIG. 3A shows a state before execution of migration.

  In FIG. 3A, as in the example shown in FIG. 2, it is assumed that the instance 22a is the migration target instance 22 in the current migration. As shown in FIG. 3A, the instance 22a belongs to the cluster a.

  As shown in FIG. 3A, the physical server 20 on which the migration target instance 22a operates is the migration source physical server 20b. The migration destination physical server 20a is the physical server 20 that is the migration destination of the migration target instance 22a.

  As shown in FIG. 3A, the migration source physical server 20b is in a state where resources are saturated by the instance 22a and the instance 22b, and a migration resource cannot be secured as it is.

  Here, assuming that there is no instance belonging to the same cluster a as the migration target instance 22a, the migration resource cannot be secured on the migration source physical server 20b by the procedure described in FIG. Therefore, in the example of the present embodiment shown in FIG. 3, the migration control unit 100 adjusts the resource allocation amount of the instance 22b operating on the migration source physical server 20b and not being migrated, thereby moving the migration source physical server 20b. Reserve migration resources on top.

  Specifically, the migration control unit 100 searches for a physical server 20 having sufficient resources in the physical server 20 in which the instance 22 belonging to the same cluster b as the instance 22b operates. Here, it is assumed that the migration control unit 100 detects the physical server 20 on which the instance 22d belonging to the same cluster b as the instance 22b operates as the physical server 20d with a margin.

  In the present embodiment, it is assumed that the instance 22b and the instance 22d belonging to the same cluster b are instances 22 that provide the same tenant service. As shown in FIG. 3A, in the load distribution of the cluster b, the SLB 30 performs load distribution so that the load of the instance 22b and the load of the instance 22d are the same during normal operation.

  FIG. 3B shows securing resources necessary for migration.

  In the example shown in FIG. 3, the migration to the migration source physical server 20b is performed by adjusting the resource allocation amount of the instance 22b on the migration source physical server 20b and the resource allocation amount of the instance 22d on the physical server 20d with margin. Resources are secured.

  The migration control unit 100 instructs the virtualization mechanism control unit 21 of the physical server 20d with a margin to increase the resource allocation amount of the instance 22d belonging to the same cluster b as the instance 22b by the migration resource amount Δ. To do. As shown in FIG. 3B, the virtualization mechanism control unit 21 of the physical server 20d having a margin increases the resource allocation amount of the instance 22d by the migration resource amount Δ.

  The migration control unit 100 instructs the SLB 30 to change the setting of the load distribution ratio between the instance 22b and the instance 22d. As shown in FIG. 3B, the SLB 30 changes the setting of the load distribution ratio between the instance 22b and the instance 22d. The changed load distribution ratio is the resource allocation amount of the original instance 22b−Δ: the resource allocation amount of the original instance 22d + Δ.

  The migration control unit 100 reduces the resource allocation amount of the instance 22b that is not the migration target by the resource amount Δ required for migration, against the virtualization mechanism control unit 21 of the migration source physical server 20b in which the resources are saturated. To instruct. As shown in FIG. 3B, the virtualization mechanism control unit 21 of the migration source physical server 20b reduces the resource allocation amount of the instance 22b that is not the migration target by the resource amount Δ required for migration.

  Migration resources are secured in the source physical server 20b by such control of the migration control unit 100. At this time, the resource reduction of the instance 22b for securing resources in the source physical server 20b is dealt with by the increase in resources of the instance 22d that performs the same service processing as the instance 22b. It does not decline.

  FIG. 3C shows execution of migration.

  The virtualization mechanism control unit 21 of the migration source physical server 20b and the virtualization mechanism control unit 21 of the migration destination physical server 20a execute migration according to an instruction from the migration control unit 100. At this time, migration resources are allocated to the migration source physical server 20b and the migration destination physical server 20a as shown in the hatched portion of FIG.

  Due to the migration, as shown in FIG. 3C, the instance 22a on the migration source physical server 20b is migrated to the migration destination physical server 20a.

  FIG. 3D shows restoration of resource allocation after the migration is completed.

  At the stage where the migration shown in FIG. 3C is completed, the resource allocation amount of the instance 22b on the migration source physical server 20b remains the original resource allocation amount -Δ, and the instance 22d on the physical server 20d with a margin. The original resource allocation amount + Δ remains the same. Although it is possible to operate the instance 22b and the instance 22d as they are, in this embodiment, the migration control unit 100 restores the resource allocation amounts of the instance 22b and the instance 22d whose resource allocation amounts have been changed.

  The migration control unit 100 causes the virtualization mechanism control unit 21 of the migration source physical server 20b to restore the resource allocation amount of the instance 22b to the original resource allocation amount, that is, the resource allocation amount of the instance 22b is Δ minutes. Instruct to increase only. As shown in FIG. 3D, the virtualization mechanism control unit 21 of the migration source physical server 20b increases the resource allocation amount of the instance 22b by Δ.

  The migration control unit 100 instructs the SLB 30 to restore the setting of the load distribution ratio between the instance 22b and the instance 22d. As shown in FIG. 3D, the SLB 30 restores the setting of the load distribution ratio between the instance 22b and the instance 22d. The restored load distribution ratio is the resource allocation amount of the original instance 22b: the resource allocation amount of the original instance 22d.

  The migration control unit 100 allows the virtualization mechanism control unit 21 of the physical server 20d with sufficient capacity to restore the resource allocation amount of the instance 22d to the original resource allocation amount, that is, the resource allocation amount of the instance 22d is Δ minutes. Instruct to reduce only. As shown in FIG. 3D, the virtualization mechanism control unit 21 of the physical server 20d with margin reduces the resource allocation amount of the instance 22d by Δ.

  By such control of the migration control unit 100, the resource allocation amount of the instance 22b and the instance 22d whose resource allocation amount has been changed to secure the migration resource is restored to the state before the occurrence of migration.

  FIG. 4 is a diagram illustrating a configuration example of the migration control unit according to the present embodiment.

  In FIG. 4, solid arrows indicate the flow of control, and broken arrows indicate the flow of data.

  The migration control unit 100 controls the movement of the instance 22 between the physical servers 20 in the data center system 1 shown in FIG. In the present embodiment, the migration control unit 100 is provided in the management mechanism device 10 shown in FIG.

  The migration control unit 100 includes a migration target / destination acquisition unit 101, a migration resource securing determination unit 102, an adjustment target instance selecting unit 103, a migration resource securing unit 106, a migration executing unit 107, and a migration resource releasing unit 108. The migration control unit 100 also includes a management information storage unit 110, a migration resource allocation range information storage unit 120, and a temporary resource allocation information storage unit 130. Each functional unit included in the migration control unit 100 is realized by hardware such as a CPU and a memory included in the computer of the management mechanism device 10 and a software program.

  The migration target / destination acquisition unit 101 acquires selection information of the migration target instance 22 and the migration destination physical server 20.

  In the present embodiment, the administrator of the data center system 1 selects the instance 22 to be moved and the physical server 20 to be moved, and inputs the selection instruction to the management mechanism device 10. The migration target / destination acquisition unit 101 acquires the instructions of the migration target instance 22 and the migration destination physical server 20 input by the administrator.

  The management mechanism control device 10 and the virtualization mechanism control unit 21 of each physical server 20 automatically select the migration target instance 22 and the migration destination physical server 20 according to the resource status of each physical server 20. You may make it do. In this case, the migration target / destination acquisition unit 101 acquires information on the migration target instance 22 and the migration destination physical server 20 selected by the management mechanism control device 10 or the virtualization mechanism control unit 21.

  Based on the resource management information stored in the management information storage unit 110, the migration resource securing determination unit 102 secures a predetermined amount of migration resources on the migration source physical server 20 and the migration destination physical server 20. Determine whether there are enough resources available. If the source physical server 20 and the destination physical server 20 have sufficient resources, the process of changing the resource allocation amount of the instance 22 as described with reference to FIGS. However, migration can be performed. The predetermined amount of migration resources is set in advance as an appropriate amount or range amount, and the setting information is stored in the migration resource allocation range information storage unit 120.

  The management information storage unit 110 stores management information of the data center system 1 such as resource management information for managing the resource status of each physical server 20 and information on a cluster in which the instances 22 are grouped. Storage device.

  FIG. 5 is a diagram showing an example of resource management data according to the present embodiment.

  5 is an example of the resource management information stored in the management information storage unit 110. The resource management data 111 illustrated in FIG. The resource management data 111 shown in FIG. 5 also includes cluster information obtained by grouping the instances 22.

  The resource management data 111 shown in FIG. 5 has information such as a physical server, a resource amount, an instance, a belonging cluster, and a resource allocation amount. In the resource management data 111 shown in FIG. 5, identification information for identifying the physical server 20 such as the name of the physical server 20 provided in the data center system 1 is recorded in the physical server. Information indicating the total amount of resources of each physical server 20 is recorded in the resource amount. In the instance, identification information for identifying the instance 22 such as the name of the instance 22 operating on each physical server 20 is recorded. In the belonging cluster, identification information for identifying the cluster such as the name of the cluster to which each instance 22 belongs is recorded. In the resource allocation amount, the resource amount allocated to each instance 22 is recorded.

  For example, in the resource management data 111 shown in FIG. 5, the resource amount of the physical server 20 of SVR # 2 is 10. On the other hand, the resource allocation amounts of the instance 22 of INS # 1 and the instance 22 of INS # 2 operating on the physical server 20 of SVR # 2 are 8 and 2, respectively. That is, since the sum of the resource allocation amounts of the instance 22 of INS # 1 and the instance 22 of INS # 2 is 10, it can be seen that the physical server 20 of SVR # 2 is in a state where resources are saturated.

  In FIG. 4, a migration resource allocation range information storage unit 120 is a storage device that can be accessed by a computer and stores setting information in which an amount or range amount considered to be appropriate as a migration resource is set.

  FIG. 6 is a diagram showing an example of migration resource allocation range data according to this embodiment.

  The migration resource allocation range data 121 illustrated in FIG. 6 is an example of setting information stored in the migration resource allocation range information storage unit 120 in which the migration resource range amount is set.

  The resource amount necessary for migration may be set assuming a constant value, but may be set assuming a range amount as shown in FIG.

  If the setting is made so that a large amount of resources is allocated as the migration resource, the instance 22 can be transferred at high speed, but it may be difficult to secure the set amount of resources. If a setting is made so that a small resource amount is allocated as a migration resource, it becomes easy to secure the set resource amount, but the transfer time of the instance 22 becomes long.

  In the present embodiment, the resource amount necessary for migration is set in a range, and migration allocation is controlled so that as much resource amount as possible is allocated to migration within the range.

In the migration resource allocation range data 121 shown in FIG. 6, delta _min represents the minimum value of the range set as a migration resources. Δ _max indicates the maximum value of the range set as the migration resource.

The delta _min, to perform the migration in a reasonable time, the amount of resources that are considered the minimum required is set. Further, Δ _max is set with a resource amount that is considered to prevent the migration speed from increasing even if more resources are allocated.

  In FIG. 4, the adjustment target instance selection unit 103 selects the instance 22 for adjusting the resource allocation amount in order to secure the migration resource.

  More specifically, the adjustment target instance selection unit 103 selects the target instance 22 to be adjusted to reduce the resource allocation amount from the instances 22 operating on the physical server 20 on which the migration target instance 22 operates. Hereinafter, the instance 22 to be adjusted to reduce the resource allocation amount is referred to as a reduced adjustment instance.

  In addition, the adjustment target instance selection unit 103 selects an instance 22 that belongs to the same cluster as the reduced adjustment instance and operates on the physical server 20 with sufficient resources as the instance 22 to be adjusted to increase the resource allocation amount. To do. Hereinafter, the instance 22 to be adjusted to increase the resource allocation amount is referred to as an increase adjustment instance. Since it is assumed that the physical server 20 on which the migration target instance 22 operates has no resources, the physical server 20 on which the increase adjustment instance operates is the physical server on which the migration target instance 22 operates. It becomes a physical server 20 other than 20.

  The adjustment target instance selecting unit 103 includes a movement target belonging cluster instance selecting unit 104 and a non-moving target belonging cluster instance selecting unit 105.

  As illustrated in the example of FIG. 2, the movement target affiliation cluster instance selection unit 104 selects the movement target instance 22 as a reduction adjustment instance. Further, the movement target affiliation cluster instance selection unit 104 selects an instance 22 that operates on the physical server 20 having sufficient resources and belongs to the same cluster as the movement target instance 22 as an increase adjustment instance.

  As illustrated in the example of FIG. 3, the non-migration target cluster instance selection unit 105 selects an instance 22 operating on the migration source physical server 20 other than the migration target instance 22 as a reduction adjustment instance. Further, the non-migration target cluster instance selection unit 105 selects, as an increase adjustment instance, an instance 22 that operates on the physical server 20 with sufficient resources and belongs to the same cluster as the decrease adjustment instance.

  The adjustment target instance selection unit 103 first attempts to select the instance 22 to be adjusted in the resource allocation amount by the movement target cluster instance selection unit 104. When the appropriate instance 22 cannot be selected by the migration target cluster instance selection unit 104, the adjustment target instance selection unit 103 uses the non-migration target cluster instance selection unit 105 to adjust the resource allocation amount of the instance 22 Try to choose.

  The adjustment target instance selection unit 103 generates information for managing the temporary adjustment status of the resource allocation based on the selected instance of the resource allocation adjustment target, and stores the information in the temporary resource allocation information storage unit 130. .

  The temporary resource allocation information storage unit 130 is a computer-accessible storage device that stores information for managing the temporary adjustment status of the resource allocation amount.

  FIG. 7 is a diagram illustrating a configuration example of temporary resource allocation data according to the present embodiment.

  The temporary resource allocation data 131 shown in FIG. 7 is an example of information for managing the temporary resource allocation amount adjustment status stored in the temporary resource allocation information storage unit 130.

  In the temporary resource allocation data 131 shown in FIG. 7, the identification information of the migration target instance 22 is recorded in the migration target instance. In the allocation change instance, the identification information of the instance 22 that is the adjustment target of the resource allocation amount, that is, the instance 22 selected as the decrease adjustment instance, and the identification information of the instance 22 selected as the increase adjustment instance are recorded.

  In the temporary resource allocation data 131 shown in FIG. 7, the pre-migration allocation server includes the identification information of the physical server 20 on which the reduced adjustment instance operates before the migration execution, and the physical server 20 on which the increased adjustment instance operates before the migration execution. Identification information is recorded. In the post-migration allocation server, identification information of the physical server 20 in which the reduced adjustment instance operates after execution of migration and identification information of the physical server 20 in which the increased adjustment instance operates after execution of migration are recorded.

  In the temporary resource allocation data 131 shown in FIG. 7, the original resource allocation amount is allocated to the reduced adjustment instance before the migration occurs, and the original resource allocation amount is allocated to the increased adjustment instance before the migration occurs. The original amount of the resource that has been recorded is recorded. The resource allocation amount during migration records the resource allocation amount of the reduced adjustment instance adjusted to secure the migration resource and the resource allocation amount of the increase adjustment instance adjusted to secure the migration resource. .

  In FIG. 4, the migration resource securing unit 106 changes the resource allocation amount of the instance 22 for which the selected resource allocation amount is to be adjusted, and allocates migration resources to the migration source physical server 20 and the migration destination physical server 20. Secure.

  More specifically, the migration resource securing unit 106 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130 and increases the resource allocation amount of the increase adjustment instance. The migration resource securing unit 106 changes the load distribution ratio between the decrease adjustment instance and the increase adjustment instance in the SLB 30. The migration resource securing unit 106 reduces the resource allocation amount of the reduced adjustment instance. The migration resource securing unit 106 secures migration resources in the migration source physical server 20 and the migration destination physical server 20. The migration resource securing unit 106 records the resource allocation amount adjustment result and the migration resource securing result in the resource management data 111 stored in the management information storage unit 110.

  The migration execution unit 107 executes migration in which the migration target instance 22 is migrated from the migration source physical server 20 to the migration destination physical server 20. The migration execution unit 107 records the migration execution result in the resource management data 111 stored in the management information storage unit 110.

  The migration resource release unit 108 releases the migration resources of the migration source physical server 20 and the migration destination physical server 20 after the migration is completed, and restores the resource allocation amount of the instance 22 whose resource allocation amount is to be adjusted. .

  More specifically, the migration resource release unit 108 releases the migration resources of the migration source physical server 20 and the migration destination physical server 20. The migration resource releasing unit 108 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130 and restores the resource allocation amount of the reduced adjustment instance, that is, uses the resource allocation amount of the reduced adjustment instance as the original resource allocation. Increase to amount. The migration resource releasing unit 108 restores the load distribution ratio between the decrease adjustment instance and the increase adjustment instance in the SLB 30. The migration resource releasing unit 108 restores the resource allocation amount of the increase adjustment instance, that is, decreases the resource allocation amount of the increase adjustment instance to the original resource allocation amount. The migration resource release unit 108 records the migration resource release result and the resource allocation restoration result in the resource management data 111 stored in the management information storage unit 110.

  Hereinafter, the flow of processing performed by the migration control unit 100 according to the present embodiment will be described with reference to FIGS.

  FIG. 8 is a migration resource securing process flowchart by the migration control unit of the present embodiment.

  In the migration control unit 100, the migration resource securing determination unit 102 performs migration resource securing determination processing using the migration source physical server 20 and the migration destination physical server 20 of the migration target instance 22 as the investigation target physical server 20. Execute (Step S10). Details of the migration resource securing determination process will be described later. The migration resource securing determination unit 102 determines whether migration resources can be secured in the migration source physical server 20 and the migration destination physical server 20 (step S11).

  If the migration resource can be secured (YES in step S11), the migration resource securing unit 106 secures the migration resource in the migration source physical server 20 and the migration destination physical server 20 (step S21).

  If the migration resource cannot be secured (NO in step S11), in the adjustment target instance selection unit 103, the migration target belonging cluster instance selection unit 104 executes a migration target belonging cluster instance selection process (step S12). Details of the movement target cluster instance selection process will be described later. The adjustment target instance selection unit 103 determines whether the instance 22 to be adjusted is selected by the movement target affiliation cluster instance selection process (step S13).

  If the instance 22 to be adjusted is selected (YES in step S13), the adjustment target instance selecting unit 103 creates temporary resource allocation data 131 based on the selection result obtained by the migration target belonging cluster instance selection process (step S13). S17). The created temporary resource allocation data 131 is stored in the temporary resource allocation information storage unit 130.

  If the instance 22 to be adjusted is not selected (NO in step S13), in the adjustment target instance selection unit 103, the cluster instance selection unit 105 that is not to be moved executes cluster instance selection processing that is not to be moved (step) S14). Details of the cluster instance selection process not to be moved will be described later. The adjustment target instance selection unit 103 determines whether or not the instance 22 to be adjusted has been selected by the non-movement target cluster instance selection process (step S15).

  If the instance 22 to be adjusted is not selected (NO in step S15), the migration control unit 100 notifies the migration resource securing failure (step S16) and ends the process.

  If the instance 22 to be adjusted is selected (YES in step S15), the adjustment target instance selecting unit 103 creates temporary resource allocation data 131 based on the selection result of the non-moving target belonging cluster instance selection process ( Step S17). The created temporary resource allocation data 131 is stored in the temporary resource allocation information storage unit 130.

  The migration resource securing unit 106 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130, and increases the resource allocation amount of the increase adjustment instance (step S18). The migration resource securing unit 106 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130, and changes the load distribution ratio between the decreasing adjustment instance and the increasing adjustment instance in the SLB 30 (step S19). . The migration resource securing unit 106 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130, and reduces the resource allocation amount of the reduced adjustment instance (step S20). The migration resource securing unit 106 secures migration resources in the migration source physical server 20 and the migration destination physical server 20 (step S21).

  FIG. 9 is a flowchart of a movement target affiliation cluster instance selection process performed by the movement target affiliation cluster instance selection unit according to this embodiment.

  The movement target affiliation cluster instance selection unit 104 initializes Δ to 0 (step S30). Further, the movement target affiliation cluster instance selection unit 104 initializes adjustment ins / svr to null (step S31).

  The migration target affiliation cluster instance selection unit 104 refers to the resource management data 111 stored in the management information storage unit 110 and selects one unprocessed instance 22 belonging to the same cluster as the migration target instance 22 (step S32).

  The movement target affiliation cluster instance selection unit 104 determines whether an unprocessed instance 22 has been selected (step S33).

  If the unprocessed instance 22 is not selected (NO in step S33), the movement target belonging cluster instance selection unit 104 determines that the process has been completed for all the instances 22, and ends the process. The value of Δ at this time is the resource amount allocated to the migration resource, and the instance 22 set in the adjustment ins / svr becomes the increase adjustment instance. The reduction adjustment instance is the instance 22 to be moved. At this time, if Δ = 0 and adjustment ins / svr = null, the instance 22 to be adjusted is not selected by the movement target cluster instance selection process.

  If an unprocessed instance 22 is selected (YES in step S33), the migration target cluster instance selection unit 104 executes a migration resource securing determination process (step S34). At this time, the migration target affiliation cluster instance selection unit 104 sets the physical server 20 on which the selected instance 22 operates and the physical server 20 of the migration destination as the physical server 20 to be investigated. The migration target affiliation cluster instance selection unit 104 substitutes the return value of the migration resource securing determination process into ret Δ. Details of the migration resource securing determination process will be described later.

  The migration target affiliation cluster instance selection unit 104 determines whether the migration resource can be secured in the result of the migration resource securing determination process (step S35).

  If the migration resource cannot be secured (NO in step S35), the migration target belonging cluster instance selection unit 104 returns to step S32 and performs processing for the next instance 22.

  If migration resources can be secured (YES in step S35), the migration target belonging cluster instance selection unit 104 determines whether ret Δ> Δ is satisfied (step S36). This determination is a determination for selecting an increase adjustment instance that allows the resource allocation amount to be adjusted so that as many resources as possible are allocated as migration resources.

  If ret Δ> Δ is not satisfied (NO in step S36), the movement target affiliation cluster instance selection unit 104 returns to step S32 and processes the next instance 22.

  If ret Δ> Δ (YES in step S36), the movement-target affiliation cluster instance selection unit 104 substitutes ret Δ for Δ (step S37). Further, the movement target affiliation cluster instance selection unit 104 sets the identification information of the current processing target instance 22 in the adjustment ins / svr (step S38). The movement target affiliation cluster instance selection unit 104 returns to step S32 and performs processing for the next instance 22.

  FIG. 10 is a flowchart of the non-move target cluster instance selection process by the non-move target cluster instance selection unit of the present embodiment.

  The cluster instance selection unit 105 not to be migrated initializes Δ to 0 (step S40). Further, the non-migration target cluster instance selection unit 105 initializes the decrease adjustment ins / svr to null, and initializes the increase adjustment ins / svr to null (step S41).

  The non-migration target cluster instance selection unit 105 refers to the resource management data 111 stored in the management information storage unit 110, and selects one unprocessed instance 22 that operates on the same physical server 20 as the migration target instance 22. Select (step S42). The unprocessed instance 22 selected in step S42 is set to ins α.

  The cluster instance selection unit 105 not to be migrated determines whether unprocessed ins α is selected (step S43).

  If unprocessed ins α is not selected (NO in step S43), the unassigned cluster instance selection unit 105 determines that the processing has been completed for all instances 22 and ends the processing. The value of Δ at this time is the resource amount allocated to the migration resource, and the instance 22 set in the decrease adjustment ins / svr becomes the decrease adjustment instance, and the instance 22 set in the increase adjustment ins / svr becomes the increase adjustment instance. . At this time, if Δ = 0, decreasing adjustment ins / svr = null, increasing adjustment ins / svr = null, the instance 22 to be adjusted has not been selected by the cluster instance selection process not to be moved. .

  If unprocessed ins α is selected (YES in step S43), the non-migration target cluster instance selection unit 105 refers to the resource management data 111 stored in the management information storage unit 110, and uses the same cluster as ins α. One unprocessed instance 22 belonging to is selected (step S44). The unprocessed instance 22 selected in step S44 is defined as ins β.

  The non-migration target cluster instance selection unit 105 determines whether an unprocessed ins β is selected (step S45).

  If unprocessed ins β is not selected (NO in step S45), the non-migration target cluster instance selection unit 105 determines that the processing has been completed for all instances 22 belonging to the same cluster as ins α. The non-migration target cluster instance selection unit 105 returns to step S42 and performs processing for the next ins α.

  If unprocessed ins β is selected (YES in step S45), the non-migration target cluster instance selection unit 105 executes a migration resource securing determination process (step S46). At this time, the non-migration target cluster instance selection unit 105 sets the physical server 20 on which ins α operates, the physical server 20 on which ins β operates, and the physical server 20 to be migrated as the physical server 20 to be investigated. To do. The non-migration target cluster instance selection unit 105 substitutes the return value of the migration resource securing determination process into ret Δ. Details of the migration resource securing determination process will be described later.

  The non-migration target cluster instance selection unit 105 determines whether the migration resource can be secured in the result of the migration resource securing determination process (step S47).

  If the migration resource cannot be secured (NO in step S47), the non-migration target cluster instance selection unit 105 returns to step S44 and processes the next ins β.

  If migration resources can be secured (YES in step S47), the non-migration target cluster instance selection unit 105 determines whether or not ret Δ> Δ (step S48). This determination is a determination for selecting a combination of a decrease adjustment instance and an increase adjustment instance that enables the resource allocation amount to be adjusted so that as many resources as possible are allocated as migration resources.

  If ret Δ> Δ is not satisfied (NO in step S48), the non-movement-target cluster instance selection unit 105 returns to step S44 and performs processing for the next ins β.

  If ret Δ> Δ (YES in step S48), the non-movement-target cluster instance selection unit 105 substitutes ret Δ for Δ (step S49). Further, the cluster instance selection unit 105 that is not to be moved sets the identification information of the current ins α in the decrease adjustment ins / svr, and sets the identification information of the current ins β in the increase adjustment ins / svr (step S50). . The non-migration target cluster instance selection unit 105 returns to step S44 and performs processing for the next ins β.

  FIG. 11 is a migration resource securing determination processing flowchart by the migration resource securing determining unit, the migration target belonging cluster instance selecting unit, and the non-moving target belonging cluster instance selecting unit according to the present embodiment.

  In the present embodiment, it is assumed that the migration resource securing determination unit 102, the migration target belonging cluster instance selecting unit 104, and the migration target nonexisting cluster instance selecting unit 105 use the same algorithm. In the following, the flowchart shown in FIG. 11 will be described as the migration resource securing determination process by the migration resource securing determining unit 102.

  The migration resource securing determination unit 102 initializes Δ to infinity (step S60).

  The migration resource securing determination unit 102 selects one unprocessed physical server 20 from the physical servers 20 to be investigated (step S61). The migration resource securing determination unit 102 determines whether an unprocessed physical server 20 to be investigated is selected (step S62).

  If an unprocessed physical server 20 to be investigated is selected (YES in step S62), the migration resource securing determination unit 102 determines that the selected physical server 20 to be investigated is the physical destination of the migration target instance 22. It is determined whether it is the server 20 (step S63).

  If the physical server 20 to be investigated is the migration destination physical server 20 (YES in step S63), the amount of resource allocation of the migration target instance 22 is subtracted from the remaining resource amount of the migration destination physical server 20 (step S63). S64), the amount of resources that can be secured as migration resources. For the physical servers 20 other than the migration destination physical server 20, the remaining resource amount is set as a resource amount that can be secured as a migration resource.

The migration resource securing determination unit 102 refers to the migration resource allocation range data 121 stored in the migration resource allocation range information storage unit 120 and determines whether the amount of resources that can be secured is Δ _min or more (step S65).

If the amount of securable resources is not equal to or greater than the delta _min (NO in step S65), the migration resource securement determination unit 102 determines that the secure migration resources fails, returning to that effect (step S66).

If the amount securable resources delta _min or more (YES in step S65), the migration resource securement determination unit 102 determines whether the amount that can be secured resources is less than delta (step S67). If the amount of resources that can be secured is less than Δ (YES in step S67), the migration resource securing determination unit 102 substitutes the amount of resources that can be secured in Δ (step S68). The migration resource securing determination unit 102 returns to step S61 and performs processing for the next physical server 20 to be investigated.

If an unprocessed physical server 20 to be investigated is not selected (NO in step S62), the migration resource securing determination unit 102 determines that the processing has been completed for all physical servers 20 to be investigated. Migration resource securement determination unit 102 refers to the migration resource allocation range data 121 stored in the migration resource allocation range information storage unit 120, determines whether the Δ> Δ _max (step S69). Migration resource securement determination unit 102, if Δ> Δ _max (YES in step S69), substitutes delta _max in delta (step S70). The migration resource securing determination unit 102 determines that the securing of the migration resource is successful, and returns Δ as a return value (step S71).

  FIG. 12 is a flowchart of migration resource release processing by the migration resource release unit of this embodiment.

  The migration resource release unit 108 releases the migration resources of the migration source physical server 20 and the migration destination physical server 20 (step S80).

  The migration resource releasing unit 108 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130 and restores the resource allocation amount of the reduced adjustment instance (step S81).

  The migration resource releasing unit 108 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130, and restores the load distribution ratio between the decrease adjustment instance and the increase adjustment instance in the SLB 30 (step S82). .

  The migration resource releasing unit 108 refers to the temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130, and restores the resource allocation amount of the increase adjustment instance (step S83).

[Example 1]
Hereinafter, a more specific example 1 of the movement of the instance 22 by the migration control unit 100 according to the present embodiment will be described with reference to FIGS. 13 to 16.

  The first embodiment is an example in which the migration resource is secured in the migration source physical server 20 by adjusting the resource allocation amount of the migration target instance 22. Here, it is assumed that the instance 22 of INS # 1 is selected as the instance 22 to be moved, and the physical server 20 of SVR # 1 is selected as the physical server 20 to be moved.

  FIG. 13 is a diagram illustrating a state before execution of migration in the first embodiment.

  FIG. 13A shows the state of each physical server 20 and SLB 30. In FIG. 13A, it is assumed that the thick and thin long frame in each physical server 20 represents the amount of resources that each physical server 20 has. Further, it is assumed that the thin line frame on the thick line frame representing the resource amount of each physical server 20 represents each instance 22 operating on each physical server 20 and its resource allocation amount. FIG. 13B shows the resource management data 111 stored in the management information storage unit 110. FIG. 13C shows temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130. The same applies to FIGS. 14 to 16.

  As shown in FIG. 13A, the physical server 20 of SVR # 1 is the destination physical server. In addition, the physical server 20 of SVR # 2 on which the instance 22 of INS # 1, which is the migration target instance, is the migration source physical server.

  As shown in FIGS. 13A and 13B, the source physical server (SVR # 2) is in a state where resources are saturated. Therefore, the migration control unit 100 adjusts the resource allocation amount for securing the migration resource. In the migration control unit 100, the adjustment target instance selection unit 103 selects the instance 22 for adjusting the resource allocation amount.

  As shown in FIG. 13B, the migration target instance (INS # 1) belongs to the cluster of CLU-a. As shown in the resource management data 111 in FIG. 13B, the physical server 20 of SVR # 3 on which the instance 22 of INS # 3 belonging to the same cluster (CLU-a) as the migration target instance (INS # 1) operates. , There is a margin of resources. The adjustment target instance selection unit 103 selects the movement target instance (INS # 1) as the de-adjustment instance, and selects the instance 22 of INS # 3 as the increase adjustment instance. As shown in FIGS. 13A and 13B, the physical server 20 of SVR # 3 on which the instance 22 of INS # 3 operates becomes a physical server with a margin.

  At this stage, in the SLB 30, the load distribution ratio between the instance 22 of INS # 1 and the instance 22 of INS # 3 in the cluster (CLU-a) is 8: 8. Also, as shown in FIG. 13C, at this stage, the temporary resource allocation data 131 is empty data.

  FIG. 14 is a diagram illustrating a state where migration resources are secured in the first embodiment.

The adjustment target instance selection unit 103 creates temporary resource allocation data 131 as shown in FIG. 14C based on the selection result of the adjustment target instance 22. Incidentally, the adjustment target instance selection unit 103, as the amount of migration resource delta, biomass 1 is ensured a maximum value delta _max migration resource allocation range.

  In the migration control unit 100, the migration resource securing unit 106 adjusts the resource allocation amount of the decrease adjustment instance (INS # 1) and the increase adjustment instance (INS # 3), and moves to the transfer destination physical server (SVR # 1) and the transfer source. Allocate migration resources to the physical server (SVR # 2).

  That is, the migration resource securing unit 106, based on the temporary resource allocation data 131 shown in FIG. 14C, increases the adjustment instance (INS) of the physical server (SVR # 3) with a margin as shown in FIG. Increase the resource allocation # 3) by Δ, that is, by 1. Based on the temporary resource allocation data 131 shown in FIG. 14C, the migration resource securing unit 106, as shown in FIG. 14A, and the INS # 1 instance 22 and INS in the cluster (CLU-a) of the SLB 30 The load distribution ratio with the instance 22 of # 3 is changed to 7: 9. Based on the temporary resource allocation data 131 shown in FIG. 14C, the migration resource securing unit 106, as shown in FIG. 14A, reduces the adjustment instance (INS # 1) of the source physical server (SVR # 2). ) Is reduced by Δ, that is, by 1. As shown in the hatched portion of FIG. 14A, the migration resource securing unit 106 assigns the resource allocation amount Δ, that is, the resource, to the migration destination physical server (SVR # 1) and the migration source physical server (SVR # 2). Allocate a migration resource with a quota of 1.

  The migration resource securing unit 106 reflects the adjustment status of these resource allocation amounts in the resource management data 111. The resource management data 111 after securing the migration resource is as shown in FIG. In FIG. 14B, a Mig resource indicates a migration resource.

  FIG. 15 is a diagram illustrating a migration execution state in the first embodiment.

  In the migration control unit 100, the migration execution unit 107 moves the migration target instance (INS # 1) from the migration source physical server (SVR # 2) to the migration destination physical server (SVR # 1) as shown in FIG. ) The instance to be moved (INS # 1) is moved while the state in which the resource allocation amount is reduced by Δ is maintained.

  The migration execution unit 107 reflects the migration execution result in the resource management data 111. The resource management data 111 immediately after the execution of migration is as shown in FIG.

  FIG. 16 is a diagram illustrating a state where migration resources are released in the first embodiment.

  In the migration control unit 100, the migration resource release unit 108 releases the migration resources of the migration destination physical server (SVR # 1) and the migration source physical server (SVR # 2), and increases the decrease adjustment instance (INS # 1). Restore the resource allocation of the adjustment instance (INS # 3).

  That is, the migration resource releasing unit 108 releases migration resources of the migration destination physical server (SVR # 1) and the migration source physical server (SVR # 2) as shown in FIG. Based on the temporary resource allocation data 131 shown in FIG. 16C, the migration resource releasing unit 108 moves the reduced adjustment instance (SVR # 1) moved to the destination physical server (SVR # 1) as shown in FIG. Increase the resource allocation of INS # 1) by Δ, that is, by 1. Based on the temporary resource allocation data 131 shown in FIG. 16C, the migration resource releasing unit 108, as shown in FIG. 16A, the INS # 1 instance 22 and INS in the cluster (CLU-a) of the SLB 30 The load distribution ratio with the instance 22 of # 3 is restored to 8 to 8. Based on the temporary resource allocation data 131 shown in FIG. 16C, the migration resource releasing unit 108, as shown in FIG. 16A, increases the adjustment instance (INS # 3) of the physical server (SVR # 3) with room. ) Is reduced by Δ, that is, by 1.

  The migration resource release unit 108 reflects the restoration status of these resource allocation amounts in the resource management data 111. The resource management data 111 after the migration resource restoration is as shown in FIG.

  The migration control unit 100 clears the temporary resource allocation data 131 shown in FIG. 16C after the series of release processing by the migration resource release unit 108 is completed.

[Example 2]
Hereinafter, a more specific example 2 of the movement of the instance 22 by the migration control unit 100 according to the present embodiment will be described with reference to FIGS.

  The second embodiment is an example in which migration resources are secured in the migration source physical server 20 by adjusting the resource allocation amount of the instances 22 other than the migration target instance 22 that operate on the migration source physical server 20. Here, it is assumed that the instance 22 of INS # 4 is selected as the instance 22 to be moved, and the physical server 20 of SVR # 4 is selected as the physical server 20 to be moved.

  FIG. 17 is a diagram illustrating a state before migration execution in the second embodiment.

  FIG. 17A shows the state of each physical server 20 and SLB 30. In FIG. 17 (A), it is assumed that the thick and thin long frame in each physical server 20 represents the amount of resources that each physical server 20 has. Further, it is assumed that the thin line frame on the thick line frame representing the resource amount of each physical server 20 represents each instance 22 operating on each physical server 20 and its resource allocation amount. FIG. 17B shows the resource management data 111 stored in the management information storage unit 110. FIG. 17C shows temporary resource allocation data 131 stored in the temporary resource allocation information storage unit 130. The same applies to FIGS. 18 to 20.

  As shown in FIG. 17A, the physical server 20 of SVR # 4 is the movement destination physical server. Further, the physical server 20 of SVR # 5 on which the instance 22 of INS # 4 that is the migration target instance operates is the source physical server.

  As shown in FIGS. 17A and 17B, the migration source physical server (SVR # 5) is in a state where resources are saturated. Therefore, the migration control unit 100 adjusts the resource allocation amount for securing the migration resource. In the migration control unit 100, the adjustment target instance selection unit 103 selects the instance 22 for adjusting the resource allocation amount.

  As shown in FIG. 17B, the migration target instance (INS # 4) belongs to the CLU-c cluster. As shown in the resource management data 111 of FIG. 17B, there is no instance 22 belonging to the same cluster (CLU-c) as the migration target instance (INS # 4). Therefore, in the second embodiment, the migration target instance (INS # 4) does not become the instance 22 whose resource allocation is to be adjusted.

  As shown in FIG. 17B, the INS # 5 instance 22 running on the same source physical server as the migration target instance (INS # 4) has a CLU-d different from the migration target instance (INS # 4). Belong to the cluster. As shown in the resource management data 111 in FIG. 17B, the resources of the SVR # 6 physical server 20 running the INS # 6 instance 22 belonging to the same cluster (CLU-d) as the INS # 5 instance 22 I can afford. The adjustment target instance selection unit 103 selects the instance 22 of INS # 5 as the decrease adjustment instance, and selects the instance 22 of INS # 6 as the increase adjustment instance. As shown in FIGS. 17A and 17B, the physical server 20 of SVR # 6 on which the instance 22 of INS # 6 operates becomes a physical server with a margin.

  At this stage, in the SLB 30, the load distribution ratio between the instance 22 of INS # 5 and the instance 22 of INS # 6 in the cluster (CLU-d) is 4 to 4. Also, as shown in FIG. 17C, at this stage, the temporary resource allocation data 131 is empty data.

  FIG. 18 is a diagram illustrating a state where migration resources are secured in the second embodiment.

The adjustment target instance selection unit 103 creates temporary resource allocation data 131 as shown in FIG. 18C based on the selection result of the adjustment target instance 22. Incidentally, the adjustment target instance selection unit 103, as the amount of migration resource delta, biomass 1 is ensured a maximum value delta _max migration resource allocation range.

  In the migration control unit 100, the migration resource securing unit 106 adjusts the resource allocation amount of the decrease adjustment instance (INS # 5) and the increase adjustment instance (INS # 6), and moves to the transfer destination physical server (SVR # 4) and the transfer source. Allocate migration resources to the physical server (SVR # 5).

  That is, the migration resource securing unit 106, based on the temporary resource allocation data 131 shown in FIG. 18 (C), as shown in FIG. 18 (A), has an increased adjustment instance (INS) of the physical server (SVR # 6) with a margin. Increase the resource allocation of # 6) by Δ, that is, by 1. Based on the temporary resource allocation data 131 shown in FIG. 18 (C), the migration resource securing unit 106, as shown in FIG. 18 (A), and the INS # 5 instance 22 and INS in the cluster (CLU-d) of the SLB 30 The load distribution ratio with the instance 22 of # 6 is changed to 3 to 5. Based on the temporary resource allocation data 131 shown in FIG. 18C, the migration resource securing unit 106, as shown in FIG. 18A, reduces the adjustment instance (INS # 5) of the source physical server (SVR # 5). ) Is reduced by Δ, that is, by 1. As shown in the hatched portion of FIG. 18A, the migration resource securing unit 106 assigns the resource allocation amount Δ, that is, the resource, to the migration destination physical server (SVR # 4) and the migration source physical server (SVR # 5). Allocate a migration resource with a quota of 1.

  The migration resource securing unit 106 reflects the adjustment status of these resource allocation amounts in the resource management data 111. The resource management data 111 after securing the migration resource is as shown in FIG. In FIG. 18B, a Mig resource indicates a migration resource.

  FIG. 19 is a diagram illustrating a migration execution state in the second embodiment.

  In the migration control unit 100, the migration execution unit 107 moves the migration target instance (INS # 4) from the migration source physical server (SVR # 5) to the migration destination physical server (SVR # 4) as shown in FIG. )

  The migration execution unit 107 reflects the migration execution result in the resource management data 111. The resource management data 111 immediately after the execution of migration is as shown in FIG.

  FIG. 20 is a diagram illustrating a state where migration resources are released in the second embodiment.

  In the migration control unit 100, the migration resource release unit 108 releases the migration resources of the migration destination physical server (SVR # 4) and the migration source physical server (SVR # 5), and increases the decrease adjustment instance (INS # 5). Restore the resource allocation of the adjustment instance (INS # 6).

  That is, the migration resource releasing unit 108 releases migration resources of the migration destination physical server (SVR # 4) and the migration source physical server (SVR # 5) as shown in FIG. Based on the temporary resource allocation data 131 shown in FIG. 20 (C), the migration resource releasing unit 108, as shown in FIG. 20 (A), the reduced adjustment instance (SVR # 5) operating on the source physical server (SVR # 5). INS # 5) resource allocation is increased by Δ, that is, by 1. Based on the temporary resource allocation data 131 shown in FIG. 20C, the migration resource releasing unit 108, as shown in FIG. 20A, the INS # 5 instance 22 and INS in the cluster (CLU-d) of the SLB 30 The load distribution ratio with the instance 22 of # 6 is restored to 4 to 4. Based on the temporary resource allocation data 131 shown in FIG. 20C, the migration resource releasing unit 108, as shown in FIG. 20A, increases the adjustment instance (INS # 6) of the physical server (SVR # 6) with a margin. ) Is reduced by Δ, that is, by 1.

  The migration resource release unit 108 reflects the restoration status of these resource allocation amounts in the resource management data 111. The resource management data 111 after the migration resource restoration is as shown in FIG.

  The migration control unit 100 clears the temporary resource allocation data 131 shown in FIG. 20C after the series of release processing by the migration resource release unit 108 is completed.

  The processing by the migration control unit 100 described above can be realized by hardware such as CPU and memory provided in the computer and a software program, and the program can be recorded on a computer-readable recording medium through the network. It is also possible to do.

  Although the present embodiment has been described above, the present invention can naturally be modified in various ways within the scope of the gist thereof.

  For example, in the present embodiment, the adjustment target instance selection unit 103 preferentially selects the movement target instance 22 as the reduction adjustment instance when the movement target instance 22 can be set as the reduction adjustment instance. Processing to do. For example, regardless of whether or not the instance 22 is the migration target instance, the instance 22 that can secure the most migration resources by adjusting the resource allocation amount among the instances 22 operating on the migration source physical server 20 It may be selected as a reduction adjustment instance. When the instance 22 to be moved is a reduced adjustment instance, the resource allocation amount of the instance 22 to be moved at the time of migration execution is reduced by adjusting the resource allocation amount, so that the migration execution time can be shortened.

DESCRIPTION OF SYMBOLS 1 Data center system 2 Tenant terminal 3 End user terminal 4 Service management network 5 Internet 10 Management mechanism apparatus 20 Physical server 21 Virtualization mechanism control part 22 Instance 30 SLB
DESCRIPTION OF SYMBOLS 40 Service network 50 Management network 100 Migration control part 101 Migration target / destination acquisition part 102 Migration resource securing determination part 103 Adjustment target instance selection part 104 Migration target affiliation cluster instance selection part 105 Non-migration affiliation cluster instance selection part 106 Migration resource securing unit 107 Migration executing unit 108 Migration resource releasing unit 110 Management information storage unit 120 Migration resource allocation range information storage unit 130 Temporary resource allocation information storage unit

Claims (5)

A program for causing a computer to control movement of a virtual machine between physical machines in a virtual machine system having a plurality of physical machines on which the virtual machine can operate,
In the computer,
Grouped by resource allocation information including resource allocation information for virtual machines for each physical machine, stored in a storage device accessible by the computer, and virtual machines performing the same processing by load balancing was based on the information, and a reduced adjusted virtual machine is a reduction target stock quota virtual machines belonging to other groups that run on the mobile object moves virtual machine that runs Domoto physical machine, wherein as the increased adjusted virtual machine is an increase target resource quotas virtual machines running reduced adjusting virtual machine and belonging to the same group and the migration source physical machine other than the physical machine, a step of respectively selecting,
Assignment by adding a predetermined amount of resources required for the movement of the up adjustments virtual machine with respect to the movement target virtual machine, have rows adjustment to reduce the predetermined resource amount from the down adjusting virtual machine, the later the adjustment virtual machine migration control program for executing the steps of securing the predetermined resource amount as a resource for moving the migration source physical machine to Oite the movement target virtual machine. The procedure for selecting the combination of the reduced adjustment virtual machine and the increase adjustment virtual machine is as follows:
A procedure for selecting the migration target virtual machine as the reduced adjustment virtual machine and selecting a virtual machine that belongs to the same group as the migration target virtual machine and that operates on a physical machine other than the migration source physical machine as the increased adjustment virtual machine When,
When the migration target virtual machine is selected as the reduced adjustment virtual machine, when a predetermined amount of resources used for virtual machine migration cannot be secured, virtual machines other than the migration target virtual machine on the migration source physical machine are selected as the reduced adjustment virtual machine. And selecting a virtual machine that operates as a physical machine other than the migration source physical machine that belongs to the same group as the reduced adjustment virtual machine and that is selected as a machine as the reduced adjustment virtual machine. The virtual machine movement control program according to 1. In the computer,
Storing temporary resource allocation information, which is information relating to adjustment of resource allocation amounts of the reduced adjustment virtual machine and the increased adjustment virtual machine, in a storage device accessible by the computer;
A procedure for restoring the resource allocation amount of the reduced adjustment virtual machine and the resource allocation amount of the increase adjustment virtual machine to the amount before adjustment of the resource allocation amount based on the temporary resource allocation information after execution of the virtual machine movement The virtual machine migration control program according to claim 1 or 2, wherein the virtual machine migration control program is further executed. In a virtual machine system having a plurality of physical machines on which a virtual machine can operate, a virtual machine movement control device that controls movement of a virtual machine between physical machines,
A storage unit for storing resource allocation information including information on a resource allocation amount for a virtual machine for each physical machine, and information obtained by grouping the virtual machines for each virtual machine performing the same processing by load balancing ;
On the basis of the resource allocation information and the virtual machine information grouped, reduced resource quotas virtual machine object belonging to another group operating on the movement target virtual machine operation to that transfer Domoto physical machine as a reduced adjustment virtual machine is, as increasing adjusted virtual machine is an increase target resource quotas virtual machines running belongs and the movement of other than the original physical machine physical machine in the same group as the reduced adjustment virtual machine, and resource allocation amount adjustment target virtual machine selection unit for respectively selecting,
Assignment by adding a predetermined amount of resources required for the movement of the up adjustments virtual machine with respect to the movement target virtual machine, have rows adjustment to reduce the predetermined resource amount from the down adjusting virtual machine, the later the adjustment virtual machine movement control apparatus comprising: a virtual machine moving resource securing unit to secure a predetermined amount of resources to the mobile based on physical machine as a resource for moving the Oite the movement target virtual machine. A virtual machine movement control method for controlling movement of a virtual machine between physical machines in a virtual machine system having a plurality of physical machines on which the virtual machine can operate.
A computer included in the virtual machine system includes:
Grouped by resource allocation information including resource allocation information for virtual machines for each physical machine, stored in a storage device accessible by the computer, and virtual machines performing the same processing by load balancing was based on the information, and a reduced adjusted virtual machine is a reduction target stock quota virtual machines belonging to other groups that run on the mobile object moves virtual machine that runs Domoto physical machine, wherein as increasing adjustment virtual machine is an increase target resource quotas virtual machines running reduced adjusting virtual machine and belonging to the same group and the migration source physical machine other than the physical machine, the steps of respectively selected,
Assignment by adding a predetermined amount of resources required for the movement of the up adjustments virtual machine with respect to the movement target virtual machine, have rows adjustment to reduce the predetermined resource amount from the down adjusting virtual machine, the later the adjustment virtual machine movement control method characterized by performing a process of securing the predetermined resource amount as a resource for moving the migration source physical machine to Oite the movement target virtual machine.

Images