JP2006323872A - Spare resource providing method and computer system for logical partition - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To deal with a problem caused by resource shortage generated by a sudden high load on a logical partition by using an inexpensive method, in a logical partition type computer. <P>SOLUTION: A hypervisor 1 includes: a resource management table 100 for managing resources allocated to the hypervisor; and an allocation/recovery means 30 for allocating/recovering the resources to/from respective logical partitions 12. Each logical partition 12 includes: a guest resource monitoring means 60 for monitoring a resource amount in use within a resource amount allocated to the logical partition; and an addition/separation means 40 for requesting addition/separation of resources to the allocation/recovery means 30 based on the monitoring result thereof. The allocation/recovery means 30 investigates the resource management table 100 according to the request from the addition/separation means 40, allocates a resource to the logical partition when a preliminary resource is present in the case of an allocation request, and recovers a resource to the resource management table as a preliminary resource in the case of a recovery request. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a logically partitioned computer system, and more particularly to a logically partitioned computer system in which a plurality of logical partitions are provided by a control program called a hypervisor and an operating system (guest OS) is operated in each logical partition.

  As a technique for operating a plurality of operating systems (guest OSs) on one physical computer, a logically partitioned computer system that creates and controls logical partitions is known.

  The logically partitioned computer system is configured by a control program that controls a logical partition called a hypervisor.

  The hypervisor is a control program serving as a core for controlling logical partitions, and performs allocation control of resources (computer resources) such as an instruction processor, a main storage device, and input / output channels.

  The hypervisor is a control program that logically separates physical resources and creates a wall between logical partitions in order to provide a logical partition. The role of the hypervisor is mainly to set a “wall” to hardware.

  In the logical partition computer system, for each logical partition, the amount of resources required when the guest OS is operating is estimated in advance, the logical partition configuration is defined according to the estimation, and when the logical partition is generated, resources are allocated according to the configuration definition.

  When the resources required for each logical partition are different during the day and night, an operation is performed in which the resource allocation between the logical partitions is systematically changed according to the operator's operation or the automatic operation schedule.

This operation method is described in detail in Patent Document 1.
Although Patent Document 1 is based on a virtual machine system, the virtual machine system has the same function as a logical partition, and the resource partitioning method is generally the same.

In addition, a pay-as-you-go billing system has appeared as a recent fee structure for computers.
This is because a lot of resources are loaded in advance, and the usage fee of the computer is determined according to the amount of resources used.

  Since resources are required in proportion to the applied load, a fee is charged only by using a computer.

Japanese Patent Laid-Open No. 6-10715

  In a conventional computer system, it is necessary to prepare in advance enough resources to withstand the load.

  If sufficient resources are not allocated and an excessive load is applied to the computer system, serious problems such as extremely slow processing of the guest OS may occur.

One phenomenon that makes this process extremely slow is generally called slowdown.
For example, there is thrashing as a slowdown phenomenon caused by a lack of main memory. Thrashing is a virtual storage system, and paging occurs frequently when there is a shortage of real storage devices.

  If this state becomes severe, the control program spends most of the processing time because of page-in / page-out processing, and the user program hardly operates.

  On the other hand, if too much resources are provided to prevent slowdown, the resources are not fully utilized and are wasted.

  For example, in a system that suddenly generates a high load due to concentration of access, such as a WWW server, many resources are required in preparation for a high load. On the other hand, an excessive resource is not used sufficiently at a low load. The problem is that the cost performance is low.

  In the computer system using the above-mentioned pay-as-you-go system, the user will be charged according to the amount of resources used, so it can be said that cost performance is good, but the installation of excessive resources is a burden for the vendor. Become.

  In addition, since it is a pay-as-you-go system, there is an aspect that the computer usage fee fluctuates.

  An object of the present invention is to solve the problems of the prior art of the above computer system in a logically partitioned computer system, enable additional allocation of necessary resources when operating the computer system, and reduce useless resources And providing a logically partitioned computer system that can be stably operated at a low cost.

In order to achieve the above object, the present invention provides a logically partitioned computer system in which one or more logical partitions are provided in one physical computer and an operating system is run in each logical partition.
As a spare resource automatic grant method for logical partitions,
In each logical partition, the amount of resources used within the allocated resource amount is monitored, and if the result of monitoring indicates that the resource is used above a predetermined threshold, Request allocation, request resource recovery if there is a surplus above a predetermined threshold for the entire resource,
In the physical computer, a method of receiving the request and allocating a spare resource to the logical partition in the case of an allocation request, and collecting a resource being allocated to the logical partition in the case of a collection request to make a spare resource Is taking.

In addition, as the logical partitioned computer system,
The physical computer is provided with resource management means for managing resources assigned to the physical computer, and assignment / recovery means for assigning or collecting resources for each logical partition,
Each logical partition requests monitoring means for monitoring the amount of resources used among the allocated resources, and requests the allocation / recovery means to add or disconnect resources based on the monitoring results of the monitoring means. Additional / separating means are provided.

  Then, the allocation / collection means inquires to the resource management means according to the request from the addition / separation means, and in the case of the allocation request, the resource is allocated to the logical partition when there is a spare resource, and in the case of the recovery request, the resource The management means is configured to collect resources as spare resources.

  Further, in the logically partitioned computer system, an external resource addition / separation unit is provided in the physical computer, and the external resource addition / separation unit is in response to the resource management unit according to an instruction to add or disconnect the resource from the operator. To add or disconnect external resources.

  Further, in the logical partition computer system, a management table is provided for each logical partition in the physical computer, and spare resource allocation availability information is set in this management table. When a resource allocation request is received from the logical partition addition / separation means, the content of the spare resource allocation availability information in the management table of the logical partition is examined, and the content indicates whether resource allocation is possible. In this configuration, spare resources are allocated, and if no resource allocation is indicated, spare resources are not allocated.

  Further, in the logical partition type computer system, when a logical partition to which a spare resource is allocated for each spare resource is set in the resource management unit, and an allocation request is received from the logical partition addition / separation unit, A configuration is adopted in which a reserved resource set to be assignable to the logical partition is assigned to the requested logical partition.

  As described above, according to the present invention, it is possible to reduce the cost by providing guest resource monitoring means for monitoring a resource shortage occurring in the guest OS and a spare resource that can be immediately allocated to any logical partition. Multiple guest OSes can be operated stably.

Hereinafter, an embodiment of a logically partitioned computer system according to the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing the configuration of a logically partitioned computer system according to an embodiment of the present invention. First, a system for automatically assigning spare resources to logical partitions according to the present invention will be described with reference to FIG.

  In FIG. 1, 11 is a hypervisor, 12 is a logical partition, 13 is an operating system (guest OS), 30 is a spare resource allocation / collection means, 60 is a guest resource monitoring means for monitoring the amount of resources used by the guest OS, and 40 is Guest resource adding / detaching means, 50 external resource adding / detaching means, 100 resource management table, 20 hardware resource, 14 guest resource, 15 spare resource, 16 recognized by hypervisor and guest OS No external resource, 17 is a hypervisor resource monitoring means for monitoring the resource management table of the hypervisor.

The logically partitioned computer system is configured by dividing a command processor 201, a main storage device 202, an input / output channel 203, and the like, which are resources of a physical computer, and generating a plurality of logical partitions 12.
Each logical partition 12 can run a guest OS 13. From the guest OS's perspective, the logical partition looks like one independent computer.
The basic structure of a logically partitioned computer system is well known and is described in US Pat. No. 4,843,451.

FIG. 2 is a diagram illustrating a resource configuration.
As described above, the computer system has an instruction processor, main memory, and input / output channels as resources, and the hypervisor controls the allocation of these resources to the logical partitions.
The guest resource 14, the spare resource 15, and the external resource 16 shown in FIG. 1 indicate some or all of the three types of resources shown in FIG.

  In a logical partition type computer system, when a system is designed, the load applied to the logical partition is estimated, and the amount of resources to be allocated to the logical partition is determined and operated.

  FIG. 3 is a flowchart for explaining the processing of the spare resource allocation / collection means of the hypervisor, FIG. 4 is a flowchart for explaining the processing of the guest OS guest resource adding / detaching means, and FIG. 6 is the guest OS guest resource monitoring. It is a flowchart explaining the process of a means.

Hereinafter, a method for allocating spare resources to the guest OS will be described with reference to these drawings.
The guest OS guest resource monitoring means 60 monitors the resources used by the guest OS at a certain time interval, and the surplus resources of the guest resources are reduced due to the occurrence of a heavy load, etc., and is monitored a certain number of times. As a result, if the surplus resource is below a certain threshold, or if the index indicating the reduction feature is below a certain threshold before the surplus resource is reduced, the guest OS is in a resource shortage state, or Then, it is detected that the resource is in a dangerous state (step 601).
For example, in the case of an instruction processor, when the instruction processor usage rate is 99% or more for a certain number of monitoring times and the response time of the online transaction is more than twice the normal time,
If it is a main storage device, multiple page-out commands issued by the operation monitoring program in the OS to the memory control program in the OS in order to detect a shortage of capacity in the real storage device and secure free space Nevertheless, since the virtual memory is accessed one after another and no free space is generated, when the OS detects a thrashing state,
For I / O channels, when the channel busy rate is 80% or more for a certain number of monitoring times,
If a resource shortage state is detected, the resource shortage is reported to the guest resource addition / detachment means 16 (step 603).
Note that the threshold for determining the shortage of resources is given by parameters, commands, etc. for each guest OS according to the business contents and the operating status of the system, and the user may be able to change settings during operation.

  When the guest resource addition / separation means 40 receives a resource shortage notification from the guest resource monitoring means (step 401), it determines the required resource amount from the state of the guest OS, and allocates / recovers a spare resource for the hypervisor. On the other hand, the resource type and amount are reported as a resource allocation request (step 402).

For example, in the case of the main storage device, it is reported as “main storage device 16 MB”.
Next, when the reserve resource allocation / collection means 30 of the hypervisor receives a resource allocation request from the guest resource addition / detachment means 40, it determines the request type from the guest OS (step 301). Step 302 and subsequent steps are executed.

In step 303, the resource management table 100 is referenced to secure the requested spare resource (step 304).
(The logical partition determination in step 302 will be described later.)
Details of the hypervisor resource management table 100 are shown in FIG.
In order to secure a spare resource, the logical partition allocation map 1005 in which the logical partition number of the currently allocated logical partition is set may be referred to.
The logical partition number is assigned in ascending order from 1. If the resource is not allocated to the logical partition, 0 is set. Therefore, a searchable resource can be found by searching the resource management table for a resource whose logical partition allocation map is 0.

  If there are not enough reserve resources that can be allocated, a report is sent to the operator who operates the guest OS and the computer (step 307).

  On the other hand, if the hypervisor has sufficient spare resources for allocation, the type and amount of resources allocated to the logical partition that reported the allocation request are reported (step 305), and the logical partition allocation map (1005) in the resource management table is reported. , 1010, 1013), the logical partition number of the assigned logical partition is registered (step 308).

  When the guest resource adding / separating means 40 receives a resource allocation notification from the hypervisor (step 401), the guest OS information (tables for managing resources) is stored in the guest OS so that the resource can be used. Change (step 406).

  With the above processing, the spare resource of the hypervisor is allocated to the logical partition, and the resource allocated by the guest OS can be used.

Next, a procedure for assigning to a logical partition for each resource type will be described in detail.
FIG. 7 is an example showing an assignment state of instruction processors.
Physical instruction processor 1, 2 (701, 702) to logical instruction processor 1, 2 (705, 706) in logical partition 1 (704), physical instruction processor 3 (703), logical instruction processor 3 (708) in logical partition 2 (707) Assigned to.

Here, when an instruction processor shortage occurs in the logical partition 2, the physical instruction processor 4 (704) of the spare resource is assigned as the logical instruction processor 4 (709) of the logical partition 2.
In this case, the logical partition allocation map of instruction processor number 4 in the resource management table is changed from 0 to 2.

FIG. 8 is an example showing the allocation state of the main storage device.
When main storage is allocated to logical partitions 1, 2, and 3 (803, 804, and 805), and a shortage of main storage occurs in logical partition 3, spare resources of main storage are allocated to main storage space (806) of logical partition 3 (802) is assigned.
In this case, in the resource management table, the logical partition allocation map of the main storage number 6 is changed from 0 to 3.

FIG. 9 is an example in which the input / output channel allocation state is shown.
Input / output channels 903 and 904 are assigned to the logical partition 1 (901), and input / output channels 905 are assigned to the logical partition 2 (902), and the input / output device 908 is accessed via the input / output device 907.
When a shortage of input / output channels occurs in logical partition 2, an input / output channel (906) is assigned to logical partition 2.

  In this case, the logical partition allocation map of channel number 4 is changed from 0 to 2 in the resource management table.

Next, an example of a method for collecting resources from a logical partition will be described.
After allocating spare resources to a logical partition due to a high load, the load on the logical partition is reduced, resulting in a surplus in the resource, and any of the results of multiple monitoring, for example, resources allocated to the logical partition If there is a surplus of 30% or more with respect to the entire resource, the guest resource monitoring means 60 detects the resource surplus (step 602) and reports the resource surplus to the guest resource adding / detaching means 40 (step 604).

  Upon receiving the resource surplus report (step 401), the guest resource adding / separating means 40 deletes the resource information from the internal table of the guest OS (step 403) and makes a resource collection request to the hypervisor (step 405). ).

Next, when the hypervisor reserve resource allocation / recovery unit receives a resource recovery request, it determines the request type from the guest OS (step 301), and in the case of a resource recovery request, executes step 306 and subsequent steps. In step 306, resources are collected from the logical partition, and thereafter, the logical partition allocation map (1005, 1010, 1013) of the corresponding entry in the resource management table is changed to 0 (step 308) and held as a spare resource for the hypervisor.
As described above, even when the resources of the logical partition are insufficient, the resources are immediately added, and a serious problem such as a slowdown of the guest OS can be avoided.

  Also, once a resource that has been granted to a logical partition becomes unnecessary, it is collected as a spare resource to the hypervisor, and is retained by the hypervisor again until it is assigned to a logical partition where a resource shortage has occurred.

Next, adding and disconnecting external resources will be described.
If there are spare resource allocation requests from a large number of logical partitions, the hypervisor may not be able to secure the requested spare resources.
In order to cope with such a situation in advance, a hypervisor resource monitoring means 17 for monitoring the reserve resource held by the hypervisor is provided, and a message is output when the remaining amount of the reserve resource falls below a certain threshold. Report to the operator.

This makes it possible to deal with additional purchase / expansion of resources to the computer.
The resource added to the physical computer is referred to as an external resource, and the process of adding and disconnecting external resources is described below.

Maintenance personnel add resources to the physical computer to make it physically usable.
Next, upon addition of an operator command, a resource addition request is sent to the external resource addition / detachment means. When the external resource addition / detachment means receives the resource addition request (501), it is added to the resource management table 100. An external resource is added as a spare resource (step 502) so that the increased resource can be managed by the hypervisor.

  In this way, even when the spare resources of the hypervisor are insufficient, the spare resources are added during operation without stopping the system.

On the other hand, in the case of disconnecting an external resource, similarly, when a resource disconnection request is sent to the external resource adding / disconnecting means 50 by an operator command input (501), the external resource adding / disconnecting means The requested resource information is deleted from the management table 100, separated from the spare resource, and made an external resource that can be removed from the physical computer.
Finally, the maintenance staff removes the resources that have become external resources from the computer.

  Next, in order to avoid the problem that a spare resource is used up in a logical partition that is important and high in priority, such as online production, in the event that a spare resource is used up in a certain logical partition, Two examples of flexible use will be described.

The first point is a method for enabling selection of whether or not a spare resource is allocated in a certain logical partition.
Whether to allocate a spare resource for each logical partition is selected at the time of initial setting or during operation.
The availability is used when determining whether or not the logical partition can be allocated when a resource allocation request is received from the guest OS.

  This determination corresponds to step 302 in FIG.

In addition, the availability information is stored in the logical partition management table 1101 that stores configuration information for each logical partition shown in FIG. 11 by providing an area (1105) for storing reserve resource allocation availability information.
The logical partition management table stores information for the hypervisor to manage the configuration of each logical partition, such as the logical partition name 1102, logical partition number 1103, main storage capacity 1104, etc. In the computer system, it is essential for the configuration.

  As shown in FIG. 13, the details of the determination of whether or not to assign to a logical partition (step 302), when a resource allocation request is received from the guest OS (step 301), refer to the logical partition management table of the requested logical partition. (Step 1301) If the reserve resource allocation value is acceptable, the process proceeds to the next processing (Step 303). If the value is not, nothing is done in response to the resource allocation request from the guest OS. The process ends (step 1304).

  The second point is a method of partitioning a spare resource and setting a logical partition that can be assigned to each partition.

As shown in FIG. 10, logical partition assignable maps 1006, 1011 and 1014 are set for each spare resource partition, and partitioning resources can be allocated only to logical partitions whose logical partition numbers exist in the map. .
This logical partition allocatable map is referred to when reserve resource allocation / collection means secures necessary reserve resources in response to an allocation request (step 304).
As shown in FIG. 12, the reserve resource is secured by setting the logical partition allocation map (1005, 1010, 1013) value of the resource management table to 0 and the logical partition allocation map (1006, 1011, 1014). This is achieved by searching for a resource in which the logical partition number of the logical partition for which an allocation request has been made exists.

  For example, in the case of securing the main memory, referring to the logical partition allocation map 1010 in FIG. 10 and searching the main storage devices that can be allocated to the logical partition 2 in order from the first entry, the main storage numbers 6 and 7 The main memory can be detected.

Similarly, when a main storage device that can be allocated to the logical partition 3 is similarly searched, a main storage device having a main storage number of 6 can be detected.
Similarly, if a main storage device that can be allocated to the logical partition 1 is similarly searched, the main storage device that can be allocated is not detected.

In this example, the request for the logical partition 1 is suppressed, and 64 MB can be set as the upper limit for the logical partition 3, so that it is possible to control to secure 64 MB of main storage resources for the logical partition 2.
With the above two methods, flexible use such as securing a large number of spare resources in important logical partitions becomes possible.

It is a block diagram which shows the structure of the logical partition type computer system by one Embodiment of this invention. It is a figure which shows the structure of a hardware resource. 6 is a flowchart for explaining processing of a spare resource allocation / collection unit of a hypervisor. It is a flowchart explaining the process of a guest OS guest resource addition / disconnection means. It is a flowchart explaining the process of the addition / separation means of the external resource of a hypervisor. It is a flowchart explaining the process of the guest resource monitoring means of guest OS. It is the figure which showed the example in which the instruction processor of a reserve resource is added to a guest. It is the figure which showed the example in which the main memory of a spare resource is added to a guest. It is the figure which showed the example in which the input / output channel of a reserve resource is added to a guest. It is the figure which showed the resource management table. It is the figure which showed the logical partition management table. 10 is a flowchart for explaining a reserve resource securing process of a spare resource allocation / collection unit of a hypervisor. It is a flowchart explaining the determination process of the logical partition of the spare resource allocation / collection means of the hypervisor.

Explanation of symbols

11 Hypervisor 12 Logical partition 13 Guest OS
14 guest resources 15 spare resources 16 external resources 17 hypervisor resource monitoring means 20 hardware resources 30 spare resource allocation / collection means 40 guest resource addition / detachment means 50 external resource addition / detachment means 60 guest resource monitoring means 100 resource management table

Claims (2)

A method for providing a spare resource for a logical partition in a logical partition computer system in which a plurality of logical partitions are provided in one physical computer and an operating system is operated in each logical partition,
In each of the logical partitions, the amount of resources used in the allocated amount of resources is monitored, and based on the monitoring result, resource allocation or collection is requested from the physical computer,
The physical computer receives the request, allocates a reserve resource to the logical partition in the case of an allocation request, and sets a resource currently allocated to the logical partition as a reserve resource in the case of a collection request. To provide spare resources for logical partitions In a computer system in which a plurality of logical partitions are provided in one physical computer and an operating system is operated in each logical partition,
The physical computer has resource management means for managing resources assigned to the physical computer, and assignment / recovery means for assigning or collecting resources for each logical partition,
Each of the logical partitions requests monitoring means for monitoring the amount of resources used among the allocated resource amounts, and requests the allocation / collection means to add or detach resources based on the monitoring result of the monitoring means. Have add / detach means,
The allocation / collection means inquires to the resource management means according to a request from the addition / separation means, allocates a resource to the logical partition when there is a reserve resource in the case of an allocation request, and A computer system, wherein a resource is collected by a resource management means as a reserve resource.

Images