JP5210730B2 - Virtual machine monitor and multiprocessor system - Google Patents

Description

  The present invention relates to a virtual machine that operates a virtual server on a physical computer by a virtual machine monitor and allocates an I / O device to the virtual server.

With the progress of semiconductor technology in recent years, a multi-core processor in which a plurality of cores are integrated on one die and a memory controller integrated processor in which a memory controller is mounted on a processor die have appeared. In order to make effective use of the computer resources accumulated in this way, there are many movements that reduce costs by consolidating processes that have been distributed to a plurality of servers in a single server. An effective means for such server aggregation is a method of operating a plurality of operating systems on one server by dividing the server. For server partitioning, a physical partitioning method that supports hardware partitioning in units of nodes or components such as processors (cores) and I / O devices, and a logical partitioning method realized by firmware called hypervisor or virtual machine monitor There is.

In the logical partitioning method, each operating system (guest OS) is executed on a logical processor provided by the virtual machine monitor, and a plurality of logical processors are mapped to physical processors by the virtual machine monitor, so that each unit is smaller than a node. Partitions can be divided. Furthermore, regarding the processor (core), one physical processor (a plurality of logical partitions)
It is also possible to execute while switching the core) in a time-sharing manner. This allows physical processors (
It is possible to create more logical partitions than the number of cores) and execute them simultaneously. As a virtual machine monitor software for the purpose of logical division, a technique as described in Patent Document 1 is representative.

However, since it is necessary to have an input / output port (path) compared to a processor and memory that are relatively easy to integrate, the number of I / O devices that are inherently difficult to integrate cannot be reduced. There is a tendency for the balance with / O devices to collapse. In order to increase the number of I / O devices, a measure such as adding slots using an I / O switch can be considered. However, there have been cases in which the I / O performance cannot be sufficiently obtained because the I / O switch increases the distance between the processor and memory and the I / O device.

Therefore, an approach has been adopted in which an I / O device that has been shared by a plurality of virtual servers is dedicated to a specific virtual server to ensure sufficient I / O performance. As a function for supporting the exclusive use of an I / O device by such a virtual server, as disclosed in Non-Patent Document 1, VT-d established by Intel Corporation is known.

On the other hand, with the progress of multi-core and the emergence of processors with integrated memory controllers,
The arrangement of resources such as processors, memories, and I / O devices tends to be unbalanced. In order to ensure performance and reliability on such an unbalanced system, resource allocation using physical location information is necessary. A conventional OS has a mechanism called affinity control for associating a specific processor with a memory, and ACPI (Advanced Configuration and Power Interface) is defined as a standard interface for acquiring physical position information in order to support this (non-patent) Reference 2). In this affinity control, resources are allocated by associating which CPU and memory the OS or application uses.

US Pat. No. 6,496,847 Intel Virtualization Technology for Directed I / O Architecture Specification, [online], Intel Corp, search on August 24, 2007, Internet <ftp://download.intel.com/technology/computing/vptech/Intel (r) _VT_for_Direct_IO.pdf> Advanced Configuration and Power Interface Specification Revision 3.0, [online], Hewlett-Packard, Intel, Microsoft, Phoenix, and Toshiba, August 24, 2007 search, Internet <http://www.acpi.info/>

However, in the conventional OS as described above, the position information of the processor and the memory is ACP.
It can be controlled using the mechanism of I. However, since the I / O device is a resource that is commonly referenced by all applications, Af using the physical location information of the I / O device
There was no concept of finity control. In fact, System R defined in ACPI
esource Affinity Table (SRAT) and System Locality Distance Information Table (
In SLIT), only the physical position information of the processor and the memory is a target, and the physical position information of the I / O device is not a management target.

On the other hand, if you try to allocate an I / O device to a specific virtual server using a virtual machine monitor, the physical location information of the I / O device is important for ensuring the performance and reliability of the virtual server. Parameter. However, the conventional ACPI-based interface has no means for acquiring I / O physical position information.

Even if the virtual machine monitor allocates appropriate resources to the virtual server, the guest OS on the virtual server cannot use the physical location information correctly.
There is a problem that the affinity control of S does not work correctly, and as a result, performance and reliability equivalent to those of a physical server cannot be secured.

The present invention has an interface for acquiring physical location information of an I / O device on a virtual machine monitor having a dedicated allocation function of an I / O device, and allocates resources to a virtual server using the acquired physical location information. The goal is to optimize according to a specified policy. Furthermore, it has an interface that appropriately converts and notifies the physical location information acquired by the virtual machine monitor to the virtual server, and is the same as when the guest OS is executed on the physical server.
It is an object to enable nity control to be executed even for a guest OS on a virtual server.

The present invention relates to a virtual machine monitor that connects one or more processors, one or more memories, and one or more I / O devices via an internal network, and allocates the processors, memory, and I / O devices to a virtual server. The virtual machine monitor includes hardware configuration information including physical location information of hardware including the processor, memory, I / O device, and network of the multiprocessor system. Based on the physical hardware information acquisition unit to be acquired, a reception unit that receives a generation request including the number of processors of the virtual server to be generated, the amount of memory, an I / O device and resource allocation policy, and the received generation request After assigning an I / O device to the virtual server, the assignment policy And a allocation unit for allocating the processor and memory to the virtual server to satisfy.

Furthermore, the virtual machine monitor further includes a notification unit that notifies the virtual server of physical location information of the processor, memory, and I / O device allocated to the virtual server.

Therefore, the present invention acquires physical location information of an I / O device and optimizes resource allocation to a virtual server using the acquired physical location information in accordance with a resource allocation policy specified in a generation request. Is possible.

Furthermore, by notifying the virtual server of the physical location information of the resources allocated by the virtual machine monitor, the same control as on the physical server can be realized on the virtual server.

  Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

FIG. 1 shows a first embodiment and a multiprocessor system (computer) to which the present invention is applied.
FIG. 3 is a block diagram showing a relationship between a virtual machine monitor and a guest OS that operate on the virtual machine monitor.

The multiprocessor system 100 employs a configuration in which one or more CPU sockets (processor packages) 110, a memory controller 130, and I / O hubs 160a and 160b are connected by an inter-module connection I / F (interface) 200. The inter-module connection I / F 200 forms an internal network of the multiprocessor system 100. Here, the CPU socket 110 and the memory controller 130, the CPU socket 110 and the I / O
Between the O hubs 160a and 160b, the memory controller 130 and the I / O hubs 160a and 160
The inter-module connection I / F 200 between b is not necessarily the same I / F. Even if they are different I / Fs, there is no difference in the following description. I / O hub 160
A generic name of a and 160b is an I / O hub 160. In addition, the memory controller 130 and the I / O
There may be a chip set equipped with the function of the O hub 160. Further, as shown in FIG. 1, the memory controller 130 may be mounted on the CPU socket 110. The CPU socket 110 has one or more CPU cores (processor cores) 120.
including. The memory controller 130 has one or more DIMMs (via the memory I / F 140).
Dual Inline Memory Module) 150 is connected. The I / O hub 160 is an I / O connection I / F
One or more I / O adapters 180a to 180f are mounted via 170, and an I / O device 190d is connected to the tip of the I / O adapters 180a to 180f. Further I / O connection I
There may be an I / O bridge or an I / O switch at the end of / F170 to constitute a multistage I / O. The I / O adapters 180a to 180f are NIC or HBA (Host Bus Ada
pter), etc., and these are generically named I / O adapter 180.

There are one or more service processors 220 on the multiprocessor system 100;
Module information acquisition I for physical position information and connection information of each module (physical component)
Collect via / F210. The service processor 220 is a multiprocessor system 1
It may be mounted externally as an option of 00, or may be mounted as one of I / O devices. Further, the function of the service processor 220 may be provided on another computer connected by a LAN or the like. The I / O device is composed of a device that inputs and outputs data.

A virtual machine monitor 300 is executed on the multiprocessor system 100, and the guest OS is divided into one or more virtual servers by dividing resources on the multiprocessor system 100.
Provided for 360a-360c. The virtual machine monitor 300 is a service processor 2
The module 20 receives the module physical configuration information and connection information collected via the physical hardware configuration information acquisition interface 320. The guest OSs 360a to 360c are collectively referred to as the guest OS 360.

The guest OS 360 allocates resources necessary as a virtual server to the logical hardware request I / F.
The virtual machine monitor 300 is notified by 350. When the guest OS 360 is started (that is, when the virtual server is started), the logical hardware request I / F 310 set by the administrator of the multiprocessor system 100 on the console 230 is used as the virtual machine monitor 300.
To notify. The virtual machine monitor 300 refers to the physical-logical hardware allocation table 310 and the like on the virtual machine monitor 300 and considers the resource allocation policy included in the logical hardware request I / F 350 and then determines the necessary physical resources. Information is obtained through the physical-logical hardware allocation interface 330. The virtual machine monitor 300 includes a logical hardware allocation processing unit 801 that allocates physical resources as logical hardware based on the acquired physical resource information and the logical hardware request I / F 310. Information on the physical resources allocated by the virtual machine monitor 300 is converted into logical hardware configuration information by the virtual machine monitor 300, and notified to the guest OS 360 via the logical hardware configuration information notification I / F 340.

In addition, the multiprocessor system 100 is connected to a console 230 that includes an input device and an output device and is used by an administrator, and the virtual machine monitor 300 and the service processor 22.
A command is given to 0, or a processing result is received from the virtual machine monitor 300 or the service processor 220 and displayed on the output device.

<Configuration of multiprocessor system>
Furthermore, the configuration of the multiprocessor system 100 in the first embodiment will be described in more detail. Four CPU sockets 110a to 110d are connected in a ring shape by an inter-module connection I / F 200. Each CPU socket 110 has a CPU core 120.
There are two CPU cores 120 in the entire multiprocessor system 100. In the following description, the eight CPU cores may be referred to as CPU core serial numbers # 0 to # 7 from the left in the figure.

Each memory controller 130 on each CPU socket 110 has four memory I / Fs 140, and four DIMMs (memory) 150 are connected to each. Here, in order to simplify the following description, it is assumed that each DIMM 150 has 1 GB of memory, 4 GB per CPU socket, and 16 GB of memory in the entire multiprocessor system. In the following description, the serial numbers of these 16 DIMMs 150 are assigned to DIMM # from the left in the figure.
Sometimes referred to as 0 to # 15.

The I / O hub 160 includes two I / O hubs 160a and 160b, each having two inter-module connection I / Fs 200. The I / O hub 160a is connected to the CPU socket 11.
0a and 110b and the I / O hub 160b are connected to the CPU sockets 110c and 110d, respectively. The I / O hub 160 has four I / O connection I / Fs 170 each,
An I / O slot 175 is provided at the end of the O connection I / F 170, and an I / O adapter 180 is connected thereto.

There are eight I / O slots 175 in the entire multiprocessor system 100. If the serial numbers of the I / O slots 175 are # 0 to 7 from the left in the figure, the I / O slot # 0 has an I / O slot 175.
/ O adapter 180a, I / O adapter 180b in I / O slot # 2, I / O adapter 180c in I / O slot # 3, and I / O adapter 18 in I / O slot # 4
0d is I / O adapter 180e in I / O slot # 5, and I / O slot # 7 is I / O adapter 180e.
/ O adapters 180f are connected to each other. In this example, nothing is connected to the I / O slot # 1 and the I / O slot # 6.

FIG. 64 is a configuration diagram showing a state in which the virtual machine monitor 300, the guest OS1 360a, the guest OS2 360b, and the guest OS3 360c are executed on the multiprocessor system 100 corresponding to the configuration of FIG. The program of the virtual machine monitor 300 exists on one of the I / O devices 190 or the ROM 201 of the service processor 200. When the multiprocessor system 300 is started, the program is stored in the memory 150 (memory 150 # 0 in this example). ) And run on one of the virtual machine CPU cores 120. The CPU core 120 that executes the virtual machine monitor 300 may be fixed, for example, a CPU core that is executed according to the operating state of the CPU core, such as a CPU core that is free at that time or a CPU core that has a low processing load. May be variable.
The guest OS 360 has a program on one of the I / O devices 190 allocated to the logical server 370 divided by the virtual machine monitor 300, and the memory 150 allocated to the 370 when the logical server 370 is started. Executed by CPU core 300 expanded above and assigned to 370. In the example of FIG. 64, the guest OS1 360a is expanded and executed on the memory 150 # 5, the guest OS2 360b is expanded on the memory 150 # 10, and the guest OS3 360c is expanded on the memory 150 # 13. In this example, the guest OS 360 is placed on a single memory module for easy understanding. However, depending on the guest OS size and memory interleave settings, it may be distributed on multiple memory modules. obtain.

2 to 6 show the physical hardware configuration information acquisition I / F 320 of the multiprocessor system 100 obtained by the virtual machine monitor 300 via the service processor 220. FIG. FIG. 2 shows a breakdown of the physical hardware configuration information acquisition I / F 320. The physical hardware configuration information acquisition I / F 320 includes a physical component configuration table 400 and an I / O adapter configuration table 450.
The inter-component distance correspondence table 500 and the physical network configuration table 550 are composed of four types of tables. These tables are generated (or updated) by inquiring about hardware resource information collected by the service processor 220 via the module information acquisition I / F 210 at a predetermined cycle or a predetermined timing. is there. Acquisition of physical hardware configuration information I / F 320 set by the virtual machine monitor 300
Is stored in the physical position information storage memory 165 or the memory 150 of the I / O hub 160 shown in FIG.

FIG. 3 shows a physical component configuration table 400 of the physical hardware configuration information acquisition I / F 320.
The structure of is shown. The physical component configuration table 400 includes a resource # 405 indicating a resource serial number, a resource type 410 indicating a resource type, a range 415 indicating a resource range corresponding to the serial number, a component # 420 indicating a serial number for each component, and a component type , The power consumption 430 of the resource specified by the serial number, and the power consumption 435 of the component for operating this resource.

In the description of the present embodiment, the component # 420 is the CPU socket 110 or the I / O.
It refers to an object that is physically inserted / extracted or turned on / off like the O hub 160, and a plurality of resources can be connected on one component # 420. In the example of this embodiment, the CPU core 120 and the DIMM 150 are connected to a component called a CPU socket 110, and the I / O adapter 180 is connected to a component called an I / O hub 160. If the memory controller 130 is different from the CPU socket 110 or if a chipset is present, these can also be independent components. The physical component configuration table 400 is a table showing the inclusion relationship between such resources and components. This indicates that the resource indicated by the range 415 is included in the component # 420. When the resource type 410 is a memory, in addition to the corresponding DIMM 150 number, a range of physical addresses is also shown. In this example, there is 1 GB per DIMM, so 16 GB from 0x0_0000_0000 to 0x0_3_FFFF_FFFF
It is divided into four according to the CPU socket to which. The power consumption [W] when each resource is in the operating state is indicated by 430, and the power consumption when the component that is the basis of the resource is in the operating state is indicated by 435. For example, it is assumed that core # 0 is operating and core # 1 is not operating. In this case, the power consumption of the resource is 20 because it is 20 per core,
The power consumption of the CPU socket of the base component is the physical component configuration table 40.
Since 0 is 80, the total 20 + 80 = 100 is the power consumption when only the core # 0 is operating. Here, in order to reduce the size of the physical component configuration table 400, a plurality of resources belonging to the same component in the range 415 are collectively shown as one entry, but different entries may be used for each resource.

Further, as the resource power consumption 430 and the component power consumption 435, data preset in the ROM 221 of the service processor 220 is used. This ROM has I / O
Information related to the performance of the component, such as the power consumption of the O adapter 180 and the bandwidth and latency between components, is stored.

FIG. 4 shows the configuration of the I / O adapter configuration table 450. In the I / O adapter configuration table 450, there are entries for all I / O adapters 180 minutes on the multiprocessor system 100.
I / O adapter # 455 indicating the serial number of the I / O adapter 180, I / O adapter 460 indicating the identifier of the I / O adapter, and I / O slot # 46 in which this I / O adapter is installed
5, adapter type 470 indicating the type of I / O adapter, power consumption 4 of this I / O adapter
It consists of 75 items. The adapter type 470 describes the type of adapter such as a network interface card (NIC) or a host bus adapter (HBA).

FIG. 5 shows the configuration of the inter-component distance correspondence table 500. The inter-component distance correspondence table 500 includes entries corresponding to all components on the multiprocessor system 100, and includes a component # 420 indicating a component identifier, a component type 425 indicating a component type, and other components and other components. It is composed of a pair-to-component distance 510 indicating the distance of the component. Distance between components 51
0 is also divided by all components, and indicates the distance from the component # 420 component to the component having the inter-component distance 510. When there are N components in the entire system, the distance between components 51
0 constitutes an N × N matrix. Usually, the distance to reach itself is considered to be zero, and therefore, it is normal that zeros are arranged in the diagonal portion of this matrix. In this embodiment, since it is assumed that the performance of the inter-module connection I / F 200 is equivalent, the distance is the number of times that the inter-module connection I / F 200 reaches the target component. This distance is treated as the physical position information of the component. For example, CPU
In FIG. 1, the distance from the socket 110a to the I / O hub 160b is the distance from the CPU socket 110a to the CPU socket 110d via the inter-module connection I / F 200 once, and from the CPU socket 110d to the I / O hub 160b. Inter-module connection I / F
Since 200 passes once, the distance between components becomes 2 in total.

When the inter-module connection I / F is not uniform, a physical network configuration table 55 described later is used.
A method of indicating the distance using the total value of the latency 570 at 0 is conceivable.

In this embodiment, since the memory 150 is directly connected to the CPU socket 110,
Since the distance between the CPU socket 110 and the memory 150 is 0, it is not included in the inter-component distance correspondence table 500, but the memory 150 is connected to the internal network (inter-module connection I / F 20).
0), the memory 150 is added to the component # 420, and the distance between the components may be set in the inter-component distance correspondence table 500.

The inter-component distance correspondence table 500 can be generated by the virtual machine monitor 300 based on the information of the physical hardware configuration information acquisition I / F 320. The service processor 220 may generate the inter-component distance correspondence table 500 based on the information of the physical hardware configuration information acquisition I / F 320 and notify the virtual machine monitor 300 of the table.

FIG. 6 shows the configuration of the physical network configuration table 550. Physical network configuration table 5
Reference numeral 50 denotes a network # 555 that indicates a serial number of connection of the inter-module connection I / F 200 with respect to all the inter-module connection I / Fs 200 on the multiprocessor system 100;
It is composed of items indicating which component 420a is connected to which component 420b, and the band 560 and latency 570 between the components 420a and 420b. In this embodiment, the network bandwidth 560 connecting the CPU sockets 110 is 6.
On the other hand, the network bandwidth 560 that connects the CPU socket 110 and the I / O hub 160 is half. The unit of the band 560 is, for example, [Gbps]
The unit of the latency 570 is, for example, [nsec].

The above is the breakdown of the physical hardware configuration information acquisition I / F 320. The service processor 220 collects these pieces of information from each resource component via the module information acquisition I / F 210. The virtual machine monitor 300 includes the service processor 220.
Acquires the physical hardware configuration information acquisition I / F 320. Physical network configuration table 5
50 or the component distance correspondence table 500, information that cannot be obtained only by the service processor 220 inquiring of the module alone is the module information acquisition I / F 210 when the multiprocessor system 100 is initialized or when the configuration is changed. And a method of automatically detecting a connection relationship using an inquiry protocol via the inter-module connection I / F 200, and information on a system management tool (existing outside the multiprocessor system 100) used by an administrator at the time of configuration change. A method is conceivable in which 220 and the like are stored in a non-volatile memory and configured based on the information.

<Virtual server configuration>
7 to 9 show the logical hardware request I / F 350 of the virtual server 1_370a and the configuration of the corresponding virtual server in the first embodiment.

FIG. 7 shows the configuration of the logical hardware request I / F 350a that requests the virtual machine monitor 300 for resources required to start the virtual server 1_370a (see FIGS. 9 and 16) when the guest OS 360 is started. . This logical hardware request I / F 350 a is set by the administrator on the console 230.

The logical hardware request I / F 350a is # 351 indicating the serial number of the virtual server, the guest name 352 indicating the identifier of the guest OS 360 operating on the virtual server, the I / O adapter 353 allocated to the virtual server, and the number of CPU cores allocated to the virtual server 354, a memory amount 355 allocated to the virtual server, a resource allocation policy 356 indicating a policy of resources allocated to the virtual server, and a priority 357 of the virtual server. Virtual server # 351 is an identifier for identifying a virtual server on the system, and does not necessarily need to be set by the requesting guest. The guest name 352 is information for identifying each guest OS 360, and includes, for example, the type of the guest OS 360 (Windows (registered trademark) / Linux, etc.). The I / O adapter 353 is a list of I / O adapters 180 requested by the guest OS. Here, I / O adapters 180a and 180c are requested. CPU core number 35
4 is the number of CPU cores required by the guest OS 360a. Here, 4 cores are required. A memory amount 355 is a memory amount required by the guest OS 360a. Here, 4 GB is required. The resource allocation policy 356 indicates a policy for allocating resources and is a key parameter in the present invention. As the resource allocation policy 356, for example, the following policy can be considered.
・ Performance priority: Choose an arrangement that shortens the distance between components * CPU-memory priority: Reduce the distance between the CPU and memory (effective when CPU memory access is high)
* CPU-I / O priority: The distance between the CPU and the I / O device is shortened (I /
(Effective when there are many O interrupts)
* I / O device-Memory priority: Reduce the distance between the I / O device and memory (effective when there are many DMA transfers from the I / O device)
* CPU-I / O device-memory priority: Each resource should be placed nearby (when balancing the overall performance of the virtual server)
-Reliability priority: Common components between virtual servers-Reduce network-Band priority: Increase effective network bandwidth-Power saving priority: Reduce overall system power consumption In the following description of the present embodiment, an example of a policy to be set is CPU-I /
O priority is given. However, the virtual server 1 (360a) is a multiprocessor system 10
Since it is the first virtual server on 0, there is a high possibility that the requested resource can be basically acquired. The priority level 357 is a parameter used to determine which virtual server is given priority when resources requested between virtual servers compete. As an example showing the priority, it is assumed here that the larger the integer is, the higher the priority is. However, in the first embodiment, it is assumed that all virtual servers have the same priority.

Further, in this embodiment, only one policy is given to the virtual server. However, a primary policy having a high priority and a secondary policy having a low priority are given, and the secondary policy is further satisfied as much as possible in the configuration satisfying the primary policy. Various configurations may be selected. In addition, an interface that assigns numerical weights to multiple policies is also conceivable. Here, for simplicity of explanation, only one policy is assumed.

FIG. 8 shows the physical-logical hardware corresponding to the result of the physical hardware assignment processing unit 801 of the virtual machine monitor 300 assigning physical resources based on the logical hardware request I / F 350a of the virtual server 1_370a shown in FIG. The structure and contents of the hardware allocation table 310 are shown. Note that the processing by which the virtual machine monitor 300 allocates hardware resources will be described in detail with reference to FIGS.

The physical-logical hardware allocation table 310 includes # 351 virtual server # 351 indicating the virtual server serial number, on / off 311 indicating the activation state of the virtual server, and usage resource # indicating the resource used by the virtual server. O adapter 312 and CPU core 313
, A memory 314, a used component # 315, and a used network # 316. In this embodiment, in accordance with the logical hardware request I / F 350a, the I / O adapter # 1
3, CPU cores # 0, 1, 2, 3, DIMM # 4, 5, 6, 7 to virtual server 1_370
assigned to a. Further, the corresponding physical address in the memory 314 of the used resource # is 4 GB from 0x1_0000_0000 (this 0x1_0000_000
0 may be referred to as the base address of the virtual server 1). In addition, 110a, 110b, and 160a are set as used component # 315 as identifiers of components on which these resources are mounted, and # 1, 5, and 6 are set as used network # 316 as network # used for connection. Is shown.

A specific procedure for generating the physical-logical H / W allocation table 310 of FIG. 8 from the logical hardware (HW) request I / F 350 of FIG. 7 will be described later.

FIG. 9 shows the logical hardware request I / F 350a and physical-logical hardware (HW).
Based on the allocation table 310, the logical hardware allocation processing unit 801 of the virtual machine monitor 300.
Shows the virtual server 1_370a allocated on the multiprocessor system 100. A portion surrounded by a dotted line in the figure indicates a resource allocated to the virtual server 1_370a among the resources in FIG.

<Operation in First Embodiment>
Hereinafter, the operation of the first embodiment will be described with reference to FIGS.

10 and FIG. 11 show the logical hardware request I / I for the next virtual server 2 after the virtual server 1_370a is assigned as shown in FIG. 9 by the logical hardware assignment processing unit 801 of the virtual machine monitor 300. F350b and 350c are shown. The logical hardware request I / F 350b and 350c are both the same I / O adapter 353, C
The number of PU cores 354 and the amount of memory 355 are requested. The difference is the resource allocation policy 356
However, in the logical hardware request I / F 350b, “CPU-memory priority” is set, whereas in the logical hardware request I / F 350c, “I / O-memory priority” is set. The following example shows how different resources are allocated due to differences in these policies.

FIGS. 12 and 13 show two types of results obtained when the logical hardware allocation processing unit 801 of the virtual machine monitor 300 allocates the logical hardware request I / F 350 of FIGS. 10 and 11 to the virtual server 2. The physical-logical hardware allocation table 310 is shown. In both cases, the I / O adapter 312 is the same as # 2, but the used CPU core 313 and the memory 314 are different. In the physical-logical hardware allocation table 310b of FIG.
5 and DIMMs # 8, 9, 10, and 11 are allocated, whereas the CPU cores # 6 and 7 and DIMMs # 0, 1, 2, and 3 are assigned in the physical-logical hardware allocation table 310c of FIG. Assigned. Other than these two methods, resource allocation methods are conceivable, but some allocation method candidates are narrowed down, such as eliminating allocation methods that are essentially unchanged.

Since the logical hardware request I / F 350b in FIG. 10 requests a CPU-memory priority policy, the CPU cores # 4 and # 5 of the CPU socket 110c and the memories # 8 to 11 connected to the CPU socket 110c are used. Is assigned to the virtual server 2, and the distance between the CPU and the memory is selected by setting the resource distance to zero. On the other hand, since the logical hardware request I / F 350c in FIG. 11 requests the I / O-memory priority policy, the I / O adapter # 2 is set so that the distance between the I / O device and the memory is minimized. Memory # 0 to # 3 that is closest to is selected. Thus, the logical hardware allocation processing unit 801 of the virtual machine monitor 300 can optimize the hardware resources allocated to the virtual server according to the requested policy.

FIG. 14 shows an inter-resource distance calculation table 600b in which the virtual machine monitor 300 calculates the distance between each resource, corresponding to the allocation table 310b shown in FIG. The inter-resource distance calculation table 600b is a virtual server number # 351 indicating a virtual server identifier (or serial number).
And a table for calculating a distance between resources in the virtual server for each virtual server number # 351. The category 601 includes CPU-memory and I corresponding to the resource allocation policy.
There are three categories: / O-memory and CPU-I / O. Each category 60
Inter-component distance 6 from resource from 602 to resource to 603 for each one
04 is calculated by the virtual machine monitor 300 according to the component # to which it belongs and the inter-component distance correspondence table 500. In this example, CPU cores # 4 and # 5 and DIMMs # 8 and 9
Since 10 and 11 are both mounted on the same component 110c, the inter-component distance 604 is zero. On the other hand, the distance between the component 160a on which the I / O adapter # 2 is mounted and the component 110c on which the DIMM # 8, 9, 10, 11 is mounted is 2 according to the inter-component distance correspondence table 500. The distance 604 is 2. Similarly, the distance between CPU cores # 4 and 5 and I / O adapter # 2 is 2
It becomes. Finally, the sum 605 of the distance between components is calculated for each category 601. In this example, the CPU-memory is 0, the I / O-memory is 2, and the CPU-I / O is 4.

FIG. 15 shows an inter-resource distance calculation table 600c corresponding to the physical-logical hardware allocation table 310c shown in FIG. Since the distance between the component 110d on which the CPU cores # 6 and # 7 are mounted and the component 110a on which the DIMMs # 0, 1, 2, and 3 are mounted is 1, the inter-component distance 604 is 1. On the other hand, since the distance between the component 160a on which the I / O adapter # 2 is mounted and the component 110a on which the DIMMs # 0, 1, 2, and 3 are mounted is 1, the inter-component distance 604 is 1. Finally CP
Since the distance between the component 110d on which the U cores # 6 and 7 are mounted and the component 160a on which the I / O adapter # 2 is mounted is 2, the inter-component distance 604 is 2.
The total inter-component distance 605 is 2 for CPU-memory, 1 for I / O-memory, CP
U-I / O is 4.

From the resource allocation candidates that have calculated the inter-resource distance calculation table 600 as described above,
Select the candidate that best meets the resource allocation policy. In the case of “CPU-memory priority” in 350b, a resource allocation with a smaller CPU-memory value of the sum 605 of the inter-component distances is selected. On the other hand, in the case of “I / O-memory priority”, resource allocation is selected in which the I / O-memory value of the sum 605 of the distance between components is small. When the physical-logical hardware allocation tables 310b and 310c are compared, it is 3 in the case of “CPU-memory priority”.
In the case of 10b, “I / O-memory priority”, the resource allocation of 310c is selected.

FIG. 16 is a configuration diagram of the virtual server 2_370b corresponding to the physical-logical hardware allocation table 310b illustrated in FIG. The virtual server 2_370b is represented as C in FIG.
The PU socket 110c, a memory connected to the PU socket 110c, and an I / O adapter 180b are included.

FIG. 17 is a configuration diagram of a virtual server corresponding to the physical-logical hardware allocation table 310c shown in FIG. The virtual server 2_370b has CPU cores # 6 and 7 on the CPU socket 110d and DIMMs # 0, 1 and 2 on the CPU socket 110a, as indicated by a dashed line in the figure.
3 and an I / O adapter 180b.

By the above example, due to the difference in the resource allocation policy of the logical hardware request I / F 350, different resources can be allocated even if the same I / O adapter, the number of CPU cores, and the memory amount are requested. The virtual server to be configured can be automatically configured.

<Modification 1>
Then, the 1st modification of 1st embodiment is shown. 18 to 20 show the logical hardware request I / F 350d of the virtual server 1_370a in the first modification example in FIG. 18, and from this logical hardware request I / F 350d, the logical hardware allocation processing unit 801 of the virtual machine monitor 300 is shown. FIG. 19 shows a physical-logical hardware allocation table 310d as a result of the allocation performed by FIG.
And the configuration of the virtual server 1_370a is shown in FIG. 1 as an I / O adapter
The request for 80a and 180c is the same as in the first embodiment, but unlike the first embodiment described above, the virtual server 1 requires two CPU cores and 8 GB of memory. In response to this request, the logical hardware allocation processing unit 801 of the virtual machine monitor 300 performs resource allocation as shown in FIGS. 19 and 20 from the CPU-I / O priority policy as the request policy.

Subsequently, after allocation of the virtual server 1 in FIGS. 21 to 28 in FIGS. 21 to 28, a resource request for the virtual server 2 is made by the logical hardware request I / F 350e shown in FIG. 21, and the logical hardware request shown in FIG. The case where the resource request | requirement of the virtual server 2 by I / F350f is performed is shown. The virtual server 2 is similar to the first embodiment in that the I / O adapter 180b and C
It requires memory with 2 or 4 GB PU cores. On the other hand, in the logical hardware request I / F 350e of FIG. 21, “CPU-memory priority” is specified as the resource allocation policy 356, and “CPU-I / O priority” is specified as the resource allocation policy 356 in 350f.

FIG. 23 and FIG. 24 show the physical-logical hardware allocation table 31 of FIG. 21 and FIG. 22, respectively.
Examples of different resource allocations for the virtual servers 2 of 0e and 310f in the virtual machine monitor 3
The logical hardware allocation processing unit 801 of 00 performs the allocation. In the physical-logical hardware allocation table 310e, the CPU core and the memory are allocated on the same CPU socket 110c, whereas in the physical-logical hardware allocation table 310f, the CPU core is on the CPU socket 110a, and the DIMM is It is allocated on the CPU socket 110d.

FIG. 25 and FIG. 26 show the results of calculating the inter-resource distance calculation tables 600e and 600f according to the respective allocations of the physical-logical hardware allocation tables 310e and 310f. In the inter-resource distance calculation table 600e in FIG. 25, the sum 605 of the inter-component distance is 0 in the CPU-memory.
, I / O-memory is 2, and CPU-I / O is 4. On the other hand, in the inter-resource distance calculation table 600f of FIG. 26, the total inter-component distance 605 is 2 for CPU-memory, 2 for I / O-memory, and 2 for CPU-I / O. As a result, in the case of “CPU-memory priority”, C
The allocation of 310e having a small sum 605 of the distance between components of the PU-memory is “CPU”.
In the case of “-I / O priority”, the allocation of 310f having the small total sum 605 of the inter-component distances of the CPU-I / O is selected.

FIG. 27 is a configuration diagram of a virtual server corresponding to the physical-logical hardware allocation table 310e of FIG. Virtual server 2_370b is CPU core # 4 on CPU socket 110c,
5, DIMM # 8, 9, 10, 11 and an I / O adapter 180 b.

FIG. 28 is a configuration diagram of a virtual server corresponding to the physical-logical hardware allocation table 310f of FIG. Virtual server 2_370b is CPU core # 0 on CPU socket 110a,
1, DIMMs # 12, 13, 14, 15 on the CPU socket 110d, and an I / O adapter 180b.

  The above is the first modification of the first embodiment.

<Modification 2>
Then, the 2nd modification of 1st embodiment is shown. 29 to 31 show the logical hardware request I / F 350g of the virtual server 1_370a in the second modification in FIG.
The logical hardware allocation table 310g is shown in FIG. 30, and the configuration of the virtual server 1_370a is shown in FIG. The I / O adapters requesting 180a and 180c are the same as in the first embodiment described above, but only one CPU core and 2 GB memory are required.
Further, since the policy is CPU-memory-I / O priority, an arrangement is selected so that the distance between the components is as close as possible. In this case, since this is the first virtual server, first, the I / O hub 160 to which the necessary I / O adapters 180a and 180c are connected.
This policy can be satisfied by selecting a CPU socket 110a close to a and assigning a CPU core and DIMM on 110a.

FIGS. 32 to 34 show three logical hardware request I / Fs 350 that differ only in the resource allocation policy 356 for the virtual server 2. Common to all, the requested I / O adapter 353 is 180b, the requested CPU core number 354 is 1, the requested memory amount 355 is 2 GB,
The priority is 5 in common. The logical hardware request I / F 350h in FIG. 32 designates “reliability priority” as the resource allocation policy 356. The logical hardware request I / F 350 i in FIG. 33 specifies “bandwidth priority” as the resource allocation policy 356. FIG.
The logical hardware request I / F 350j of No. 4 specifies “power saving priority” as the resource allocation policy 356. In the following, we will see what configuration is selected according to each policy.

35 to 36 show the physical-logical hardware allocation table 3 corresponding to the two resource allocations.
10 is shown. In the physical-logical hardware allocation table 310h of FIG. 35, C on a CPU socket 110b different from that used by the virtual server 1 as a CPU core and memory.
PU core # 2 and DIMM # 4, 5 are allocated to satisfy the requested reliability priority policy. On the other hand, in the physical-logical hardware allocation table 310i of FIG. 36, the CPU core # 1 and DIMM # 2 on the same CPU socket 110a used by the virtual server 1 are used.
3 is satisfied, and the requested bandwidth priority policy is satisfied.

When the allocation policy is not distance priority as in the above-described embodiment, another index is required instead of the inter-resource distance calculation table 600 as a resource allocation reference.

37 to 38 show a component network allocation table 650 in place of the inter-resource distance calculation table 600. FIG. The component / network allocation table 650 is a table for the logical hardware allocation processing unit 801 of the virtual machine monitor 300 to check the components and networks shared among a plurality of virtual servers and the effective network bandwidth, and the identifier of the virtual server. Alternatively, a virtual server # 351 indicating a serial number, a shared component # 651 indicating an identifier of a component shared between virtual servers, and a shared network # indicating an identifier of a network used to use the shared component 6
52 items and items such as network # 653, shared number 654, and effective bandwidth 655 corresponding to all the networks used in each virtual server.

The component network allocation table 650h in FIG. 37 is a table corresponding to the resource allocation in the physical-logical hardware allocation table 310h in FIG. Shared component # 651
Corresponds to the I / O hub 160a, and the shared network # 652 is not provided. Since it is not shared for each network # 653, the shared number 654 is 1. Effective bandwidth 655
Is a value obtained by dividing the value of the bandwidth 560 corresponding to each network # 555 in the physical network configuration table 550 by the number of shares 654 (that is, the number of virtual servers). In this case, since the number of shares is 1, the value of the network bandwidth 560 becomes the effective bandwidth 655 as it is.

The component network allocation table 650i in FIG. 38 is a table corresponding to the resource allocation in the physical-logical hardware allocation table 310i in FIG. Shared component # 651
Are the I / O hub 160a and the CPU socket 110a. The shared network # 652 corresponds to the network # 5 that connects the I / O hub 160a and the CPU socket 110a. In this case, since the sharing number 654 is 2, the effective bandwidth 655 is half the bandwidth 560 of the original network.

If the resource allocation policy 356 is “Reliability priority”, the number of shared components
Select a configuration that minimizes the number of shared networks. In this case, since the I / O adapters requested by the virtual server 1 and the virtual server 2 are on the same I / O hub 160a, it is impossible to set the number of shared components to zero. However, the component network allocation table 650h corresponding to the physical-logical hardware allocation tables 310h and 310i, respectively.
, 650i, it can be seen that the number of shared components is smaller corresponding to the physical-logical hardware allocation table 310h shown in FIG. So the requested policy is "
For the request of the logical hardware request I / F 350h of FIG. 32 which is “reliability priority”, the allocation of the physical-logical hardware allocation table 310h shown in FIG. 35 is selected.

When the resource allocation policy 356 is “bandwidth priority”, a configuration is selected so that the effective bandwidth of the network is as large as possible. In this case, when the component network allocation tables 650h and 650i in FIG.
The effective bandwidth is larger in the configuration of 0h. Therefore, for the request of the logical hardware request I / F 350i whose policy is “bandwidth priority”, the allocation in the physical-logical hardware allocation table 310h of FIG. 37 that maximizes the effective bandwidth is selected.

FIG. 39 shows an allocation resource power consumption calculation table 700h corresponding to the physical-logical hardware allocation table 310h of FIG. 35 that is used when the requested policy is power saving priority. This table is configured by the virtual machine monitor 300 for resources and components used in the entire system, not in units of virtual servers, and shows a category 701 indicating whether the target hardware is a resource or a component, and a resource type / component type. Show 7
02, used resource / component # 703 indicating an identifier of a used resource or component, power consumption 704 indicating power consumption for each resource, and total power consumption 705
Consists of As resources, there are items of CPU core, memory, I / O slot, and I / O adapter, and used resources are listed. The power consumption 704 refers to the items of the resource power consumption 430 in the physical component configuration table 400 and the power consumption 475 in the I / O adapter configuration table 450, and sets each value. Category 701
In the component entry, the components required when using each resource are listed, and the component power consumption 704 is set according to the component power consumption 435 of the physical component configuration table 400. The total power consumption 705 is the power consumption 7
The sum of 04 is calculated by the virtual machine monitor 300.

FIG. 40 shows an allocation resource power consumption calculation table 700i corresponding to the physical-logical hardware allocation table 310i of FIG. Since the number of used resources does not change in the physical-logical hardware allocation tables 310h and 310i, the items of resources are almost the same although the numbers # of the used resources are different. The difference is in the item of the component. In the physical-logical hardware allocation table 310h, the total number of components used is three, but in the physical-logical hardware allocation table 310i, the total number of components used is two. . Therefore, when the total power consumption 705 is compared, the value is 345 [W] in the allocated resource power consumption calculation table 700h of FIG. 39, whereas it is 265 [W] in 700i.

When the resource allocation policy 356 of the logical hardware request I / F 350 is “power saving priority”, the virtual machine monitor 300 selects an allocation method such that the total power consumption 705 is as small as possible. Therefore, in response to the request of the logical hardware request I / F 350j of FIG. 34 whose policy is “power saving priority”, the logical hardware allocation processing unit 801 of the virtual machine monitor 300 performs the physical-logical hardware allocation of FIG. Select an assignment in table 310i.

FIG. 41 is a block diagram of a virtual server corresponding to the physical-logical hardware allocation table 310h of FIG.

FIG. 42 is a configuration diagram of a virtual server corresponding to the physical-logical hardware configuration table 310i of FIG.

  The above is the second modification of the first embodiment.

Note that in the case of the above policy that prioritizes power consumption, the amount of heat generated by each resource and component may be used instead of power consumption.

<Specific Resource Allocation Procedure in First Embodiment>
Here, from the logical hardware configuration request I / F 350 in the first embodiment, the physical-
A procedure for generating the logical hardware configuration table 310 will be described in detail with reference to FIGS.

FIG. 55 shows a flowchart of the entire logical hardware allocation processing performed by the logical hardware allocation processing unit 801 of the virtual machine monitor 300. First, the requested I / O adapter is allocated to the guest OS 360, and further selected and allocated from unallocated CPU and memory, evaluated according to the requested policy, and the CPU and CPU closest to the most requested policy are allocated.
The procedure is to select a combination of memories.

First, the process proceeds to step 800. The logical hardware allocation processing unit 801 of the virtual machine monitor 300 reserves an entry for the virtual server 351 in the physical-logical hardware allocation table 310. Proceed to step 805.

Step 805 determines whether the allocation request of the I / O adapter 353 requested by the logical hardware request I / F 350a is exclusive or shared. If the result of the determination is exclusive, step 81
If 0, if shared, proceed to step 830.

Step 810 determines whether the requested I / O adapter 353 can be assigned. If it is already assigned to another virtual server and cannot be assigned, the process proceeds to step 820. If it can be assigned, the process proceeds to step 830.

Step 820 is a step of responding an error to the logical hardware request I / F 350. After the error response, the process proceeds to step 890.

Step 830 is an I / O adapter allocation process. The subroutine flow of the I / O adapter assignment processing 830 is shown in FIG. After completion of this subroutine, step 8
Proceed to 40.

In step 840, the allocation candidate CPU 841 and the allocation candidate memory 842 are emptied (
This is the step set in (ii). Here, the allocation candidate CPU 841 and the allocation candidate memory 842 indicate CPU and memory combination candidates that are closest to the allocation policy among the combinations so far. At this point, no CPU or memory has been selected, so an empty set (
Ф). Proceed to step 850.

Step 850 is a process of selecting a CPU / memory satisfying the request from unallocated CPU / memory. The subroutine flow of the CPU / memory selection processing 850 is shown in FIG.
Shown in If there is an error response at the completion of this subroutine, the process proceeds to step 820. Otherwise, go to step 860.

In step 860, the CPU / memory selected in step 850 is evaluated according to the allocation policy. The flow of the policy evaluation 860 subroutine is shown in FIG. After completion of this subroutine, the routine proceeds to step 870.

Step 870 is a step of determining if there are more unassigned CPU / memory combinations. If it still remains, return to step 850. Otherwise, go to step 880.

Step 880 is processing for allocating the allocation candidate CPU 841 and the allocation candidate memory 842 to the virtual server 351 of the physical-logical hardware allocation table 310. The subroutine flow of the CPU / memory allocation process 880 is shown in FIG. When this subroutine is completed, the logical hardware allocation process is completed.

Step 890 is a step of deleting the entry of the virtual server 351 in the physical-logical hardware allocation table 310 when an error occurs during the allocation in step 850 (or when there are not enough resources allocated in step 810). is there. After this entry is deleted, the logical hardware allocation process is completed.

FIG. 56 shows a flow of a subroutine of the I / O adapter assignment processing 830. First, the process proceeds to step 900.

The step 900 selects the I / O adapter # 455 corresponding to the I / O adapter 353 of the logical hardware request I / F 350a from the I / O adapter configuration table 450, and the virtual server 351 of the physical-logical hardware configuration table 310. This is a step of adding the entry to the used I / O adapter 312. Proceed to step 901.

In step 901, the I / O slot # 465 corresponding to the I / O adapter 353 is set to I / O.
The resource type 41 of the physical component configuration table 400 is selected from the adapter configuration table 450
0 searches for the entry corresponding to I / O slot # 465 from the entry of I / O slot,
The corresponding component 420 is replaced with the virtual server 35 in the physical-logical hardware configuration table 310.
It is a step of adding to the use component 315 of one entry. Proceed to step 902.

Step 902 is a step of determining whether all the I / O adapters 353 have been allocated. If adapters that have not yet been assigned remain, the process returns to step 900. Otherwise, the I / O adapter assignment process 830 is complete.

FIG. 57 shows a flow of a subroutine of the CPU / memory selection process 850. First, the process proceeds to step 905.

In step 905, the number of CPU cores requested by the logical hardware request I / F 350a is 3
54 and whether or not the memory amount 355 can be allocated. If assignment is impossible, the process proceeds to step 907. If assignment is possible, the process proceeds to step 906.

In step 906, the logical hardware request I / F3 is selected from the unassigned combinations.
In this step, CPU cores and memories satisfying the CPU core number 354 and the memory amount 355 requested in 50a are selected and set in the temporary allocation CPU 851 and the temporary allocation memory 852. Thereby, the CPU / memory selection processing 850 is completed.

Step 907 is an error response step. As a result, the CPU / memory selection process 850 is completed.

FIG. 58 shows a flow of a subroutine of policy evaluation 860. First, the process proceeds to step 910.

Step 910 is a step of performing conditional branching according to the allocation policy 356 of the logical hardware request I / F 350a. Allocation policy 356 is CPU-memory priority,
If any of CPU-I / O priority, I / O-memory priority, and CPU-I / O-memory priority is selected, the process proceeds to step 920. If policy 356 is trustworthy, step 9
Proceed to 30. If the policy 356 has bandwidth priority, the process proceeds to step 940. When the policy 356 has priority on power saving, the process proceeds to step 950.

Step 920 is a distance calculation process between components. FIG. 59 shows a subroutine flow. After completion of this subroutine, the process proceeds to step 911.

Step 930 is a component sharing number calculation process. FIG. 60 shows a subroutine flow. Proceed to step 911 after completion.

Step 940 is an effective bandwidth calculation process. FIG. 61 shows a subroutine flow. Proceed to step 911 after completion.

Step 950 is a power consumption calculation process. FIG. 62 shows a subroutine flow. Proceed to step 911 after completion.

Step 911 is a step of determining whether the allocation candidate CPU 841 and the allocation candidate memory 842 are empty sets (Ф). If it is an empty set (Ф), the process proceeds to step 913. Otherwise, go to step 912.

Step 912 is a step of determining whether or not the temporary allocation policy value 853 is smaller than the allocation candidate policy value 843. Here, the policy value is an index for quantitatively evaluating the configuration of the allocated resource component according to each allocation policy 356, and it is defined that the smaller the value, the closer to the requested policy. If the temporary allocation policy value 853 is smaller than the allocation candidate policy value 843, the process proceeds to step 913. Otherwise, policy evaluation 860 is complete.

Step 913 includes temporary allocation CPU 851, temporary allocation memory 852, temporary allocation policy value 85.
3, assignment candidate CPU 841, assignment candidate memory 842, assignment candidate policy value 84, respectively.
This is a step of substituting into 3. By this process, the allocation candidate CPU 841, the allocation candidate memory 842, and the allocation candidate policy value 843 hold the CPU / memory combination having the smallest policy value allocated so far. With this process, the policy evaluation 860 is completed.

58, the allocation candidate policy value 843 is set so as to inherit the value of the temporary allocation policy value 853. In the first step 912, the initial value of the allocation candidate policy value 843 is indefinite. In the first time, the allocation candidate CPU 841 and the allocation candidate memory 842 are set to に お い て in step 840 of FIG. 55, so the determination in step 911 always goes to step 913, and step 912 does not pass. The allocation candidate policy value 843 may be initialized with a predetermined maximum value.

FIG. 59 shows a flow of a subroutine of the inter-component distance calculation 920. First, the process proceeds to step 921.

Step 921 is the I / O of the virtual server 351 of the logical hardware request I / F 350a.
In this step, the inter-resource distance calculation table 600 is created by using the adapter 353, the temporary allocation CPU 851, and the temporary allocation memory 852. In particular,
(1) The component to which each resource belongs is obtained from the physical component configuration table 400.
(2) The distance between each component is set to the inter-component distance 604 in the inter-resource distance calculation table 600 according to the inter-component distance correspondence table 500.
(3) The sum of the inter-component distances 604 is obtained for each category 601 and substituted into Σ605.
It consists of the process. Next, the process proceeds to step 922.

Step 922 is a step of performing conditional branching according to the allocation policy 356 of the logical hardware request I / F 350a. If the policy 356 is CPU-memory priority, the process proceeds to step 923. If policy 356 has CPU-I / O priority, step 92
Proceed to 4. If the policy 356 is I / O-memory priority, go to step 925. If the policy 356 has CPU-I / O-memory priority, the process proceeds to step 926.

Step 923 is a step in which Σ 605 corresponding to the CPU-memory of category 601 in the inter-resource distance calculation table 600 is set as the temporary allocation policy value 853. Thereby, the inter-component distance calculation 920 is completed.

Step 924 is a step in which Σ 605 corresponding to the CPU-I / O of category 601 in the inter-resource distance calculation table 600 is set as the temporary allocation policy value 853. Thereby, the inter-component distance calculation 920 is completed.

Step 925 is a step in which Σ 605 corresponding to the I / O-memory of category 601 in the inter-resource distance calculation table 600 is set as the temporary allocation policy value 853. Thereby, the inter-component distance calculation 920 is completed.

Step 926 is a step in which the sum of Σ 605 of all categories 601 in the inter-resource distance calculation table 600 is set as the temporary allocation policy value 853. Thereby, the inter-component distance calculation 920 is completed.

FIG. 60 shows a flow of a subroutine for the component sharing number calculation 930. First, the process proceeds to step 931.

Step 931 includes an I / O adapter 353 for the logical hardware request I / F 350a,
This is a step of creating the component / network allocation table 650 for all the allocated virtual servers including the virtual server 351 using the temporary allocation CPU 851 and the temporary allocation memory 852. In particular,
(1) The component to which each resource belongs is obtained from the physical component configuration table 400.
(2) A shared component # 651 is a component shared between different virtual servers.
Set to.
(3) Set the number of shared components # to the shared number 654.
It consists of the process. Proceed to step 932.
In step 932, the shared number 654 corresponding to the virtual server 351 in the component / network allocation table 650 is set as the temporary allocation policy value 853. Thereby, the component sharing number calculation 930 is completed.

FIG. 61 shows a flow of a subroutine for effective bandwidth calculation 940. First, the process proceeds to step 941.

Step 941 includes the I / O adapter 353 of the logical hardware request I / F 350a,
This is a step of creating the component / network allocation table 650 for all the allocated virtual servers including the virtual server 351 using the temporary allocation CPU 851 and the temporary allocation memory 852. In particular,
(1) The component to which each resource belongs is obtained from the physical component configuration table 400.
(2) A network to be used between components is obtained from the physical network configuration table 550 and set to the network # 653.
(3) Network # 653 shared network # 65 shared between different virtual servers
Set to 2.
(4) Obtain the bandwidth of each network from the bandwidth 560 of the physical network configuration table 550,
A value obtained by dividing this by the sharing number 654 is set to the effective bandwidth 655.
Such processing is included. Proceed to step 942.

Step 942 is a step in which the minus of the effective bandwidth 655 corresponding to the virtual server 351 in the component / network allocation table 650 is set as the temporary allocation policy value 853. Here, it is defined that the smaller the policy value 853, the closer to the requested policy. On the other hand, the larger the absolute value of the effective band, the better. In view of this, by setting a value obtained by subtracting the effective band value as the temporary allocation policy value 853, the configuration having the largest effective band is finally selected. Thereby, the execution bandwidth calculation 940 is completed.

FIG. 62 shows a flow of a subroutine of power consumption calculation 950. First, the process proceeds to step 951.

Step 951 includes a logical hardware request I / F 350a I / O adapter 353,
This is a step of calculating the allocation resource power consumption calculation table 700 for all allocated virtual servers including the virtual server 351 using the temporary allocation CPU 851 and the temporary allocation memory 852. In particular,
(1) The component to which each resource belongs is obtained from the physical component configuration table 400.
(2) From the physical component configuration table 400, the resource power consumption 430 of all allocated resources and the power consumption are obtained for the components, and the power consumption 70 of the calculation table 700 is calculated.
Set to 4.
(3) The sum of all the power consumption 704 is calculated and set to Σ705.
Such processing is included. Proceed to step 952.

Step 952 is a step in which the total power consumption 705 of the allocation resource power consumption calculation table 700 is set as the temporary allocation policy value 853. Thereby, the power consumption calculation 950 is completed.

FIG. 63 shows a flow of a subroutine of the CPU / memory allocation process 880. First, the process proceeds to step 960.

Step 960 is a step of adding the allocation candidate CPU 841 to the used CPU core 313 of the entry of the virtual server 351 in the physical-logical hardware configuration table 310 and the allocation candidate memory 842 to the used memory 314. Proceed to step 961.

In step 961, an entry corresponding to the CPU core 313 is searched from the entry of the CPU core with the resource type 410 in the physical component configuration table 400, and the corresponding component 420 is used as the entry of the virtual server 351 in the physical-logical hardware configuration table 310. Similarly, the resource type 410 is added to the component 315 and the memory type is changed from the memory entry
14 to the use component 315 of the entry corresponding to 14. Proceed to step 962.

Step 962 is a step of determining whether all allocation candidate CPUs 841 and allocation candidate memory 842 have been allocated. If the unassigned allocation candidate CPU 841 or the allocation candidate memory 842 still remains, the process returns to step 960. Otherwise, the CPU / memory allocation process 880 is complete.

Through the series of operations described above, the virtual machine monitor 300 can create the physical-logical hardware allocation table 310 from the logical hardware request I / F 350 instructed from the console 230.

As described above, according to the first embodiment, the virtual machine monitor 300 acquires the distance between components indicating the physical position information of the components in the multiprocessor system 100, and the inter-component distance correspondence table 500 is obtained. The physical component configuration table 400, the I / O adapter configuration table 450, and the physical network configuration table are set in advance by setting the physical network configuration and the power consumption of each resource. The virtual machine monitor 300 receives a logical hardware request I / F 350 that is a virtual server generation (or setting) request from the console 230, and first requests the requested I / O adapter 180.
Is selected, and the I / O adapter 180 is exclusively allocated to the requested virtual server. Next, the virtual machine monitor 300 refers to the physical location information and power consumption of the resource from the inter-component distance correspondence table 500 and the physical component configuration table 400 to the virtual server, and from the I / O device that has performed the dedicated allocation. By selecting a CPU and memory that satisfy the required policy and assigning them to the virtual server, it is possible to provide an optimal virtual server for the requested policy.

In particular, when a plurality of virtual servers (guest OSs) are operated on one physical computer, even if the policy differs for each virtual server, the physical location information of the I / O device, CPU socket, and memory is assigned to each virtual server. An optimum configuration can be automatically assigned, and the availability of the virtual machine can be improved.

Second Embodiment (Logical Hardware Configuration Notification I / F)
43 to 48 show a second embodiment of the present invention. The configuration of the multiprocessor system 100 of the second embodiment is the same as that of the first embodiment.

FIG. 43 shows the logical hardware request I / F 350k of the virtual server 1_370a. In the logical hardware request I / F 350k, four CPU cores, 8 GB of memory, I / F
180d and 180f are requested as the O adapter.

FIG. 44 shows a physical-logical hardware allocation table 310k corresponding to the logical hardware request I / F 350k of FIG. 45 shows the multiprocessor system 100 shown in FIG.
The arrangement of the virtual server 1_370a assigned above is shown. A portion surrounded by a dotted line in the figure is the virtual server 1_370a.

Here, consider the operation of the guest OS1_360a on the virtual server 1_370a. The guest OS1_360a is an OS on the multiprocessor system 100, and has an affinity control function related to the CPU and memory conforming to the ACPI as described above. Af
The CPU cores # 4, 5 and DIMMs # 8, 9, 10, 11 on the CPU socket 110c are assigned to the applications on the guest OS 360 by the finity control function, or the CPU cores # 6, 7 on the CPU socket 110d are assigned. And DIMM # 12, 13, 14, 1
By assigning 5, the latency of access from the CPU core to the memory is shortened.
Improve performance. However, in order to use such an affinity control function, the guest OS1_
360 a needs to know the hardware arrangement of the virtual server 1. Therefore, a configuration information notification I / F (logical hardware configuration information notification I / F 340) from the virtual machine monitor 300 to the guest OS 360 is required.

FIG. 46 shows physical hardware configuration information 750 on the virtual machine monitor 300. This is information that is secondarily created from the physical-logical hardware allocation table 310, host CPU core # 751 indicating the identifier or serial number of the CPU core 120, and host physical indicating the start point of the physical address of the allocated memory 150. Address base 752 and allocated memory 1
It consists of a host physical address range 753 indicating a quantity of 50. However, even if this information is directly notified to the guest OS 1_360a, the guest OS 1_360a cannot correctly perform the affinity control. This is because the host physical address base on the multiprocessor system 100 assigned by the virtual machine monitor 300 to the guest OS 360a and the guest OS
This is because it differs from the guest physical address base in 1_360a.

FIG. 48 shows the relationship between the host physical address space and the guest physical address space. When the guest OS1_360a is assigned to the host physical address base 0x2_0000_0000, an address shift occurs so that the guest physical address base 0x0_0000_0000 corresponds to the host physical address base 0x2_0000_0000. Therefore, when notifying the logical hardware configuration information to the guest OS1_360a, it is necessary to take into account this deviation.

FIG. 47 shows the logical hardware configuration information notification I / F 340 from the virtual machine monitor 300 to the guest OS1_360a. The logical hardware configuration information notification I / F 340 is sent to the guest CPU core # 341 indicating the identifier or serial number of the CPU core 120 allocated to the guest OS 360, the guest physical address base 342 indicating the base address allocated to the guest OS 360, and the guest OS 360. It consists of a guest physical address range 343 indicating the allocated address range. In FIG. 47, the guest CPU core # 341 is 0 in the guest OS 360.
Re-install in order. The guest physical address base 342 is the host physical address base 7
The value obtained by subtracting the base address 0x2_0000_0000 of the virtual server 1_370a from 52. The guest physical address range 343 has the same value as the host physical address range 753. As described above, the used logical hardware configuration information that can be used by the guest OS1_360a for the affinity control based on ACPI is created. The virtual machine monitor 300 generates this logical hardware configuration information notification I / F 340 and notifies the guest OS 360 of it.
Here, only the combination of the CPU # 341 and the memory address range is notified, but it is also possible to notify more advanced information (for example, including distance information such as the distance calculation table 600 between resources). is there. Even in this case, the principle of converting the host physical address into the guest physical address remains the same.

As described above, according to the second embodiment of the present invention, the virtual machine monitor 300 allocates appropriate resources to the virtual server, and the guest OS 360 on the virtual server has the logical hardware configuration generated by the virtual machine monitor 300. By acquiring the information notification I / F 340, it is possible to correctly use the physical location information of the component to be used, and the guest OS 3
It is possible to ensure the performance and reliability equivalent to the physical server by correctly operating the 60 affinity control.

<Third Embodiment> (Resource Reassignment)
A third embodiment of the present invention will be described with reference to FIGS. 43 to 45 and FIGS. 49 to 54.
As in the second embodiment, in accordance with the logical hardware request I / F 350k shown in FIG. 43, the virtual machine monitor 300 assigns the result of assigning the virtual server 1_370a as shown in the physical-logical hardware assignment table 310k shown in FIG. As shown in FIG.

Assume that the logical hardware request I / F 350m related to the virtual server 2 shown in FIG. 49 is input from the console 230 to the virtual machine monitor 300 in a state where the virtual server 1_370a is assigned as shown in FIG. In the logical hardware request I / F 350m of FIG. 49, the priority 357 is 10, and the logical hardware request I / F 35 of the virtual server 1 is set.
The value is larger than the priority 5 of 0k. In this case, the virtual server 2 with higher priority
The virtual machine monitor 300 is required to rearrange the resources already allocated to the virtual server 1 so as to satisfy the resource allocation policy.

FIG. 50 shows a physical-logical hardware allocation table 310m when the virtual machine monitor 300 allocates resources to the virtual server 2_370b while leaving the resources allocated to the virtual server 1_370a as they are. Further, FIG. 51 shows a physical-logical hardware allocation table 310n when resources already allocated from the virtual server 1_370a are picked up once, resources are allocated to the virtual server 2_370b, and resources are allocated again to the virtual server 1_370a. . 52 shows the physical-logical hardware allocation table 310m of FIG.
FIG. 53 shows an inter-resource distance calculation table 600n calculated according to the physical-logical hardware allocation table 310n of FIG.

The resource allocation policy 356 of the logical hardware request I / F 350m in FIG.
When the sum 605 of the distances between the components of “I / O-memory” in the resource distance calculation table 600 is compared, it is 2 in the calculation table 600m of FIG. It is 1 in the table 600n, and it can be seen that the arrangement that satisfies the resource allocation policy of the virtual server 2 is possible when the resources are rearranged. FIG. 54 shows a state in which the virtual machine monitor 300 allocates the virtual server 1_370a and the virtual server 2_370b on the multiprocessor system 100 according to the physical-logical hardware allocation table 310n.

As described above, when there are a plurality of logical hardware requests having different priorities, the resources are preferentially assigned to the requests having higher priorities by assigning the resources in order from the logical hardware request I / F 350 having higher priorities. Can be assigned. However, there are cases where it is necessary to move the CPU and memory when rearranging resources. Regarding the movement of the CPU, copying of the contents of the register, flushing of the cache and TLB, and the like are required, and regarding the movement of the memory, copying of the memory is required. A known or well-known method may be applied as a specific method for moving the CPU or memory.

As described above, according to the third embodiment of the present invention, when a virtual server is newly allocated to the multiprocessor system 100, the already allocated resources are once released, and then the virtual server has the highest priority. By reallocating resources, a more optimal virtual server can be configured.

As described above, the present invention can be applied to a computer system configured by a plurality of processors and I / O devices and divided into a plurality of virtual servers and its virtual machine monitor.

The block diagram which shows 1st Embodiment and shows the structure of the multiprocessor system and virtual machine monitor to which this invention is applied. Explanatory drawing which shows 1st Embodiment and shows the structure of physical hardware configuration information acquisition I / F. Explanatory drawing which shows 1st Embodiment and shows the structure of a physical component structure table. Explanatory drawing which shows 1st Embodiment and shows the structure of an I / O adapter structure table. Explanatory drawing which shows 1st Embodiment and shows the structure of the distance correspondence table between components. Explanatory drawing which shows 1st Embodiment and shows the structure of a physical network structure table | surface. Explanatory drawing which shows 1st Embodiment and shows the structure of the logical hardware request | requirement I / F of the virtual server 1. FIG. Explanatory drawing which shows 1st Embodiment and shows the structure of the physical-logical hardware allocation table | surface of the virtual server 1. FIG. The block diagram which shows 1st Embodiment and shows an example of the virtual server 1 on a multiprocessor system. FIG. 3 is an explanatory diagram illustrating a configuration of a logical hardware request I / F of the virtual server 2 according to the first embodiment in which the policy is CPU-memory priority. FIG. 3 is an explanatory diagram illustrating a configuration of a logical hardware request I / F of the virtual server 2 according to the first embodiment in which the policy is I / O-memory priority. Explanatory drawing which shows 1st Embodiment and shows the structure of the physical-logical hardware allocation table of the virtual server 2 which made the policy CPU-memory priority. FIG. 3 is an explanatory diagram illustrating a configuration of a logical hardware request I / F of the virtual server 2 according to the first embodiment in which the policy is I / O-memory priority. Explanatory drawing which shows 1st Embodiment and shows the structure of the resource distance calculation table | surface of the virtual server 2 which made the policy CPU-memory priority. Explanatory drawing which shows 1st Embodiment and shows the structure of the inter-resource distance calculation table | surface of the virtual server 2 which made the policy I / O-memory priority. The block diagram which shows 1st Embodiment and shows the structure of the virtual servers 1 and 2 on the multiprocessor system which made the policy CPU-memory priority. The block diagram which shows 1st Embodiment and shows the structure of the virtual servers 1 and 2 on the multiprocessor system which made policy a priority with I / O-memory. Explanatory drawing which shows a 1st modification and shows the structure of logic hardware request | requirement I / F. Explanatory drawing which shows a 1st modification and shows the structure of a physical-logical hardware allocation table. The block diagram which shows the 1st modification and shows the structure of the virtual server 1 on a multiprocessor system. Explanatory drawing which shows a 1st modification and shows the structure of the logical hardware request | requirement I / F of the virtual server 2 which made the policy CPU-memory priority. Explanatory drawing which shows a 1st modification and shows the structure of the logical hardware request | requirement I / F of the virtual server 2 which made the policy CPU-I / O priority. Explanatory drawing which shows a 1st modification and shows the structure of the physical-logical hardware allocation table of the virtual server 2 which made the policy CPU-memory priority. Explanatory drawing which shows a 1st modification and shows the structure of the physical-logical hardware allocation table | surface of the virtual server 2 which made the policy CPU-I / O priority. Explanatory drawing which shows the structure of the distance calculation table | surface of the virtual server 2 which shows the 1st modification and made the policy CPU-memory priority. Explanatory drawing which shows the structure of the distance calculation table between resources of the virtual server 2 which shows the 1st modification and made the policy CPU-I / O priority. The block diagram of the virtual servers 1 and 2 on a multiprocessor system at the time of showing a 1st modification and making a policy CPU-memory priority. The block diagram of the virtual servers 1 and 2 on a multiprocessor system at the time of showing a 1st modification and making a policy CPU-I / O priority. Explanatory drawing which shows a 2nd modification and shows the structure of the logical hardware request | requirement I / F of the virtual server 1. FIG. Explanatory drawing which shows a 2nd modification and shows the structure of the physical-logical hardware allocation table | surface of the virtual server 1. FIG. The block diagram which shows the 2nd modification and shows the structure of the virtual server 1 on a multiprocessor system. Explanatory drawing which shows the structure of the logical hardware request | requirement I / F of the virtual server 2 which showed the 2nd modification and made the policy priority on reliability. Explanatory drawing which shows the structure of the logical hardware request | requirement I / F of the virtual server 2 which showed the 2nd modification and gave priority to the policy band. Explanatory drawing which shows the structure of the logical hardware request | requirement I / F of the virtual server 2 which showed the 2nd modification and made the policy a power-saving priority. Explanatory drawing which shows the structure of the physical-logical hardware allocation table | surface of the virtual servers 1 and 2 which showed the 2nd modification and made the policy priority on reliability. Explanatory drawing which shows the structure of the physical-logical hardware allocation table | surface of the virtual servers 1 and 2 which showed the 2nd modification and gave priority to the policy band. Explanatory drawing which shows the structure of the component network allocation table | surface of the virtual servers 1 and 2 which showed the 2nd modification and made the policy priority on reliability. Explanatory drawing which shows the structure of the component network allocation table | surface of the virtual servers 1 and 2 which showed the 2nd modification and made the policy a bandwidth priority. Explanatory drawing which shows a 2nd modification and shows the structure of the allocation resource power consumption calculation table | surface at the time of making a policy priority on reliability. Explanatory drawing which shows a 2nd modification and shows the structure of the component network allocation table | surface at the time of setting a policy as power consumption priority. The block diagram on the multiprocessor system of the virtual servers 1 and 2 which showed the 2nd modification and made the policy priority on reliability. The block diagram on the multiprocessor system of the virtual servers 1 and 2 which showed the 2nd modification and gave priority to the band policy. Explanatory drawing which shows 2nd Embodiment and shows the structure of the logical hardware request | requirement I / F of the virtual server 1. FIG. Explanatory drawing which shows 2nd Embodiment and shows the structure of the physical-logical hardware allocation table | surface of the virtual server 1. FIG. The block diagram which shows 2nd Embodiment and shows the structure of the virtual server 1 on a multiprocessor system. Explanatory drawing which shows 2nd Embodiment and shows the structure of physical hardware configuration information. Explanatory drawing which shows 2nd Embodiment and shows the structure of logical hardware configuration information notification I / F. The map which shows 2nd Embodiment and shows the relationship between a host physical address and a guest physical address. Explanatory drawing which shows 3rd Embodiment and shows the structure of the logical hardware request | requirement I / F of the virtual server 2. FIG. Explanatory drawing which shows 2nd Embodiment and shows the structure of the physical-logical hardware allocation table | surface of the virtual servers 1 and 2. FIG. Explanatory drawing which shows 2nd Embodiment and shows the other structure of the physical-logical hardware allocation table | surface of the virtual servers 1 and 2. FIG. Explanatory drawing which shows 2nd Embodiment and shows the structure of the resource power consumption calculation table | surface of the virtual server 2. FIG. Explanatory drawing which shows 2nd Embodiment and shows the other structure of the resource power consumption calculation table | surface of the virtual server 2. FIG. The block diagram which shows 2nd Embodiment and shows the virtual servers 1 and 2 on a multiprocessor system. The flowchart which shows 1st Embodiment and shows an example of the logical hardware allocation process performed with a virtual machine monitor. The flowchart which shows 1st Embodiment and shows the subroutine of the I / O adapter allocation process of a logical hardware allocation process similarly. The flowchart which shows 1st Embodiment and shows the subroutine of CPU and memory selection processing of logical hardware allocation processing similarly. The flowchart which shows 1st Embodiment and shows the subroutine of the policy evaluation process of a logical hardware allocation process similarly. The flowchart which shows 1st Embodiment and similarly shows the subroutine of the distance calculation process between components of a logical hardware allocation process. The flowchart which shows 1st Embodiment and shows the subroutine of the component shared number calculation process of a logical hardware allocation process similarly. The flowchart which shows 1st Embodiment and shows the subroutine of the effective bandwidth calculation process of a logical hardware allocation process similarly. The flowchart which shows 1st Embodiment and similarly shows the subroutine of the power consumption calculation process of a logical hardware allocation process. The flowchart which shows 1st Embodiment and similarly shows the CPU of a logic hardware allocation process, and the subroutine of a memory allocation process. 1 is a configuration diagram of a multiprocessor system according to a first embodiment.

Explanation of symbols

100 multiprocessor system 110 CPU socket 120 CPU core 130 memory controller 140 memory interface 150 memory 160 I / O hub 170 I / O connection interface 180 I / O adapter 190 I / O device 200 inter-module connection interface 210 module information acquisition interface 220 Service processor 300 Virtual machine monitor 310 Physical-logical hardware allocation table 320 Physical hardware configuration information acquisition interface 330 Physical-logical hardware allocation interface 340 Logical hardware configuration information notification interface 350 Logical hardware request interface 400 Physical component configuration table 500 Inter-component distance correspondence table 550 Physical network configuration table 600 Scan distance calculation table 650 component network allocation table 700 allocated resource consumption based calculation table 750 physical HW configuration information 801 logical hardware allocation processing unit

Claims (12)

A multiprocessor comprising computer resources including a plurality of processors, a plurality of memories, and a plurality of I / O devices, wherein the computer resources are divided into groups and belong to components, and the components are connected to each other via an internal network. A virtual machine monitor that runs a virtual server on the system,
The virtual machine monitor is
Said affiliation information, and the physical hardware information acquiring unit that acquires the hardware configuration information including the information indicating the physical connections between the components,
A reception unit that receives a generation request including information on the number of processors and the amount of memory required for the virtual server to be generated, information on designation of an I / O device, and a resource allocation policy;
The I / O device specified by the received generation request is allocated to the virtual server , and the required number of processors and the required amount of memory are allocated to the received policy, the belonging information, and the physical An allocation index that is assigned to the virtual server based on an evaluation index obtained from information indicating a secure connection relationship ;
A virtual machine monitor characterized by comprising: The generation request further includes priority information of the virtual server,
The allocation processing unit
If the priority of the new virtual server requested to be generated is higher than the allocated virtual server, the allocation of the processor and memory already allocated to the allocated virtual server is once released, and the allocation is released. 2. The virtual machine monitor according to claim 1, wherein a processor and a memory to be allocated to the new virtual server are selected from a processor group including a new processor and a memory group including a memory whose allocation has been released. The physical hardware information acquisition unit obtains distances between the processor and the memory and between the processor and the I / O device from the information indicating the belonging and the physical connection relation as the evaluation index ,
The allocation processing unit
Allocating a processor and memory to the virtual server so that a total sum of the distances of the processor, memory, and I / O devices is minimized when allocating to the virtual server based on the policy and the evaluation index. The virtual machine monitor according to claim 1. The generation request further includes information on whether or not the specified I / O device is dedicated to the virtual server,
The allocation processing unit
When the generation request includes information that the specified I / O device is exclusively used by the virtual server, the specified I / O device is already assigned to another virtual server. Determine whether or not
2. The virtual machine monitor according to claim 1, wherein when the virtual server is already assigned to another virtual server, generation of the virtual server requested to be generated is canceled. The physical hardware information acquisition unit obtains a distance between a processor and a memory from the information indicating the belonging and the physical connection relationship as the evaluation index ,
The allocation processing unit
When the resource allocation policy gives priority to processor access to memory, the total distance between the processor and memory is minimized when allocating to the virtual server based on the policy and the evaluation index. The virtual machine monitor according to claim 1, wherein a processor and a memory are allocated to the virtual server. The physical hardware information acquisition unit obtains a distance between the processor and the I / O device from the information indicating the belonging and the information indicating the physical connection relationship as the evaluation index ,
The allocation processing unit
When the resource allocation policy prioritizes access between the processor and the I / O device, the total distance between the processor and the I / O device when allocating to the virtual server based on the policy and the evaluation index. The virtual machine monitor according to claim 1, wherein a processor and a memory are allocated to the virtual server such that the virtual machine is minimized. The physical hardware information acquisition unit obtains the distance between the memory and the I / O device from the affiliation information and the information indicating the physical connection relationship as the evaluation index ,
The allocation processing unit
When the resource allocation policy gives priority to access between the memory and the I / O device, the total distance between the memory and the I / O device is allocated to the virtual server based on the policy and the evaluation index. The virtual machine monitor according to claim 1, wherein a processor and a memory are allocated to the virtual server such that the virtual machine is minimized. The physical hardware information acquisition unit obtains the mutual distance between the processor, the memory, and the I / O device from the information indicating the belonging and the information indicating the physical connection relationship as the evaluation index ,
The allocation processing unit
When the resource allocation policy secures the overall performance of the virtual server, when allocating to the virtual server based on the policy and the evaluation index , the distance between the processor, the memory, and the I / O device is determined. The virtual machine monitor according to claim 1, wherein a processor and a memory are allocated to the virtual server so that the sum is minimized. The physical hardware information acquisition unit obtains a path of an internal network connecting a processor, a memory, and an I / O device from the information indicating the belonging and the information indicating the physical connection relationship as the evaluation index ,
The allocation processing unit
When the resource allocation policy gives priority to the reliability of a virtual server, when the resource is allocated to the virtual server based on the policy and the evaluation index , the internal network used by the already allocated virtual server, and the generation 2. A processor and a memory are allocated to the new virtual server so that the total number of overlapping internal networks is minimized in the internal network used by the requested new virtual server. Virtual machine monitor as described in. The physical hardware information acquisition unit obtains a path of an internal network connecting a processor, a memory, and an I / O device from the information indicating the belonging and the information indicating the physical connection relationship as the evaluation index ,
The allocation processing unit
If the resource allocation policy gives priority to the bandwidth of the internal network, the internal network used by the already allocated virtual server and the generation request when allocating to the virtual server based on the policy and the evaluation index In the internal network used by the new virtual server, the value obtained by dividing the overlapping internal network bandwidth by the number of virtual servers is defined as the effective bandwidth, and the effective bandwidth is maximized. The virtual machine monitor according to claim 1, wherein a processor and a memory are allocated. The physical hardware information acquisition unit uses the processor and memory, the I / O device, the processor, the memory, and the information based on the affiliation information and the information indicating the physical connection relationship as the evaluation index. Obtain the power consumption of the component to which each I / O device belongs,
The allocation processing unit
When the resource allocation policy prioritizes power consumption, the processor, memory, I / O device, and component of the already allocated virtual server when allocating to the virtual server based on the policy and the evaluation index Processor and memory to the new virtual server so that the total power consumption of the processor, memory, I / O devices, and components used by the new virtual server requested to generate is minimized. The virtual machine monitor according to claim 1, wherein:   The virtual machine monitor according to claim 1, wherein the component is a processor socket or an I / O hub.

Images