JP6752889B2 - Management method and management calculator - Google Patents

Description

The present invention relates to a management method.

In computer system operation management, the performance of system components such as servers and storage systems is monitored, and if performance degradation is detected, it does not matter whether the performance recovers naturally or not, and performance degradation will occur in the future. It is necessary to take measures such as configuration changes to prevent recurrence. However, there are a large number of possible combinations of the configuration changing means and the parameters of the configuration changing means. Therefore, in order to reduce the burden on the administrator, it is important to support the proposal of countermeasures to prevent the reproduction of performance degradation.

According to Patent Document 1, when the amount of I / O for the logical volume in the storage system changes and the required performance is not satisfied, the configuration information, the performance information, and the capacity consumption at the time when the required performance is not satisfied are satisfied. Based on the tendency, it is possible to generate a configuration change proposal that satisfies the required performance of the logical volume at a future point and display the most appropriate configuration change proposal. In addition, based on the information of the future configuration change schedule, it is possible to predict the future capacity consumption and the configuration change of the storage system, and display only the configuration change proposal that can be implemented at the scheduled configuration change implementation time.

International Publication No. 2014/073045

The above-mentioned Patent Document 1 can satisfy the required performance based on the configuration information and the performance information at the time when the predetermined required performance of the logical volume is not satisfied and at the time when the configuration change is completed in consideration of the time required for the configuration change. Generate possible configuration change proposals.

However, compared to the gradual change in capacity consumption tendency, the performance changes greatly from moment to moment depending on the workload of the application. Therefore, in the countermeasure plan generated by using the information only at a specific time point based on Patent Document 1, when the load at the time of performance deterioration occurs again, even if the recurrence of the performance deterioration is suppressed at the time of focus, other than that. At this point, an appropriate performance improvement effect cannot be expected and performance deterioration may occur.

On the other hand, as described above, the performance changes from moment to moment, so even if a countermeasure plan is generated based on Patent Document 1 for a plurality of past time points in which performance degradation is desired to be prevented, the countermeasure plan generated differs depending on the time of interest. there is a possibility. It takes a huge amount of calculation time to generate appropriate countermeasures at all points in time.

In order to solve the above problems, the management method according to one aspect of the present invention has a period of the bottleneck resource that is the cause of the performance deterioration of the computer system in the past period among a plurality of resources in the computer system. The bottleneck load data, which is the time-series data of the load of the bottleneck resource, is acquired, and the load of the related resource during the period is relative to the related resource that affects the load of the bottleneck resource among the plurality of resources. The performance is selected from a plurality of countermeasure policy candidates that acquire the related load data, which is the time-series data of the above, and indicate the policy for determining the parameters necessary for operating the computer system based on the tendency of the bottleneck load data. A countermeasure for the deterioration is selected, and based on the countermeasure policy and the tendency of the related load data, the countermeasure for the performance deterioration is selected from a plurality of countermeasure candidate candidates each indicating the operation of the computer system. It is provided that the means is selected and the parameters of the countermeasure means are determined based on the countermeasure policy.

Appropriate countermeasures can be quickly generated over a period of time.

An outline of the operation of the management calculator according to the first embodiment is shown. The hardware configuration of the computer system according to the first embodiment is shown. The hardware configuration of the management computer according to the first embodiment is shown. The logical configuration of the computer system according to the first embodiment is shown. The VM management table 1101 is shown. The host management table 1102 is shown. The storage volume management table 1103 is shown. The threshold management table 1104 is shown. The performance management table 1105 is shown. Countermeasure means management table 1106 is shown. The metric management table 1107 is shown. Countermeasure means application management table 1108 is shown. The selection means holding table 1109 is shown. It is a flowchart of performance monitoring processing. It is a flowchart of a countermeasure plan generation process. It is a flowchart of Script1. It is a flowchart of Script2. It is a flowchart of Script3. It is a flowchart of Script4. It is a flowchart which shows the measure policy determination process (S2020). It is a flowchart which shows the countermeasure means selection process (S2030). It is a flowchart which shows the countermeasure means selection process (S2030) which concerns on Example 2. FIG. It is a flowchart which shows the countermeasure means combination generation processing which concerns on Example 2. FIG.

Some embodiments will be described with reference to the drawings. It should be noted that the examples described below do not limit the invention according to the claims, and that all the elements and combinations thereof described in the examples are indispensable for the means for solving the invention. Not exclusively.

In the following description, various information may be described by the expression of "aaa table", but various information may be expressed by a data structure other than the table. The "aaa table" can be called "aaa information" to show that it does not depend on the data structure. Further, an information element consisting of the values of each column in the table is referred to as a column or entry, and an entry in the "aaa table" is referred to as an "aaa table entry" for the sake of explanation.

Further, in the following description, the processing may be described simply with the management computer and the server as the subjects, but these processings are executed by the processor (for example, CPU (Central Processing Unit)) of the control device included in the computer. Indicates that Similarly, when the process is simply described with the storage device as the subject, it indicates that the controller provided in the storage device is executing. Further, at least one of the control device and the controller may be the processor itself, or may include a hardware circuit that performs a part or all of the processing performed by the control device or the controller.

The program may be installed on each computer or storage device from the program source. The program source may be, for example, a program distribution server or storage media.

In the following description, operations such as configuration change operations (for example, moving a VM, changing ownership) and setting change operations (for example, setting an upper limit on the amount of I / O) are described as countermeasures. Describe the set value that needs to be entered when implementing the countermeasure measures as a parameter. In addition, the combination of countermeasure means and parameters is described as a countermeasure plan or plan.

(1-1) Outline of the embodiment of this embodiment

FIG. 1 shows an outline of the operation of the management calculator according to the first embodiment.

This figure shows an example of a method for instructing generation of a performance deterioration countermeasure plan in the computer system 10, a presentation example of a performance deterioration countermeasure plan, and a flow of generating a countermeasure plan.

The management computer 100 is connected to the computer system 10 and the display device 400. The computer system 10 includes a server 200 and a storage device 300. Further, the VM 700 may be running on the server 200. These details will be described later. The display device 400 may include an input device or may be a terminal device. The computer system may be any one of the server 200 and the storage device 300, or may be one computer that provides the VM and the volume.

As an example for showing the outline of this embodiment, FIG. 1 shows a case where the performance of the volume 1 in the storage device 300 is deteriorated and the required performance is lower than the required performance. Hereinafter, the flow of processing in this embodiment will be described.

First, when the management computer 100 detects a deterioration in the performance of the computer system, it issues an alert to the administrator. Upon receiving the alert, the administrator designates an arbitrary period in the past as a specified period, and starts generating a countermeasure plan aiming at the performance deterioration not recurring even if the load of the specified period occurs again. The method of designating the designated period may be, for example, a method of inputting the start time and end time of the designated period on the screen displayed on the display device 400, or a method of pointing a point on the graph showing the load change on the graph. A method such as moving a line segment representing the time may be used.

Further, when the management computer 100 detects a performance deterioration, a designated period may be set and the generation of the countermeasure plan may be automatically started.

The management computer 100 includes a countermeasure policy determination module, a countermeasure means selection module, and a countermeasure plan generation module. When the management computer 100 is instructed by the display device 400 to take measures against the performance deterioration, the countermeasure policy determination module uses the components of the computer system in which the performance deterioration has occurred (hereinafter, the components are used as resources). The load information, which is the tendency of the load, is analyzed by referring to the load information in the specified period (the resource that causes the performance deterioration is called the bottleneck resource). Then, the countermeasure policy judgment module assumes that the performance degradation of the bottleneck resource is reproduced based on the analysis result, and the countermeasure policy generation policy that can efficiently eliminate the performance degradation (hereinafter, simple). To be described as a countermeasure policy). In the example of this figure, the bottleneck resource is the volume 1 of the storage device 300, and the load of the bottleneck resource is represented by the operating rate.

Next, the countermeasure selection module is the cause of increasing the load of the bottleneck resource (affecting the load of the bottleneck resource), and is a component of the computer system (hereinafter referred to as a related resource for simplicity). ) Refers to the load information in the specified period and analyzes the load tendency of each related resource. Then, based on the analysis result, only the countermeasure measures effective in eliminating the performance deterioration are selected based on the countermeasure policy decided in advance. In the example of this figure, the related resources are VM1 and VM2 of the server 200, and the load of the related resources is represented by the amount of I / O.

After that, the countermeasure plan generation module determines the calculation method of the parameters of the countermeasure means determined earlier. Then, the parameter (set value of the countermeasure means) is determined based on the determined calculation method.

In the example of this figure, the countermeasure policy determination module selects to uniformly reduce the load at each time point as the countermeasure policy based on the bottleneck resource load tendency. The countermeasure means selection module selects the storage side I / O amount limit as the countermeasure means based on the countermeasure policy and the related resource load tendency. The countermeasure generation module determines the amount of I / O to be limited.

The management computer 100 causes the display device 400 to display the determined countermeasure plan. The display method may be, for example, a method of displaying the countermeasure means (Method) and its parameters (Detail) in text. In addition, the displayed contents may include elements other than the countermeasure means and parameters, for example, additional information such as an outline of the countermeasure plan, and a plurality of countermeasure plans may be displayed.

Further, the countermeasure plan is not only displayed on the display device 400, but may be saved as electronic data such as a file.

(1-2) Hardware configuration of computer system

FIG. 2 shows the hardware configuration of the computer system according to the first embodiment.

The server 200 is connected to the management computer 100, the storage device 300, and the display device 400 via a first communication network 500 such as a LAN (Local Area Network). Further, the server 200 is connected to the storage device 300 via a second communication network 600 such as a SAN (Storage Area Network). The first communication network 500 and the second communication network 600 may be integrally formed.

The management computer 100 includes an I / F 130 such as a NIC (Network Interface Card). Details will be described later as a description of FIG.

The server 200 includes a processor 210, a storage resource 220, a management I / F (hereinafter referred to as M-I / F) 230, and a communication I / F (hereinafter referred to as C-I / F) 240. It is a physical computer that is composed.

The MI / F 230 is a communication interface device for communicating with the first protocol, for example, a NIC. The C-I / F240 is an interface device for communicating with the second protocol, and is, for example, an HBA (Host Bus Adapter).

The storage resource 220 is, for example, a memory, and may include an auxiliary storage device such as an HDD (Hard Disk Drive). The storage resource 220 stores an application program such as an OS (Operating System) or a business application, and the processor 210 executes the application program or the OS.

The storage device 300 has a storage device group 360 and a controller 390 connected to the storage device group 360. The storage device group 360 is composed of one or more storage devices. The storage device is, for example, an SSD (Solid State Drive), and may be another non-volatile memory such as an HDD or FeRAM. As described above, the storage device 300 may include a plurality of storage devices having different performances. The storage device group 360 may be provided from the outside of the storage device 300.

The controller 390 has a memory 320, an MI / F 330, a CI / F 340, a device I / F (DI / F) 350, a cache memory 380, and a processor 310 connected to them. ..

The DI / F350 is a communication interface device for communicating with a storage device by a third protocol. The DI / F 350 may be provided for each type of storage device.

The memory 320 stores a computer program executed by the processor 310 and various information. Further, the cache memory 380 temporarily stores data according to a write command received from the server 200 and data read from the storage device in response to a read command received from the server 200. The memory 320 and the cache memory 380 may be an integrated device.

The display device 400 is a device including, for example, an input / output device such as a display, a keyboard, and a pointer device, but other devices may be used. In addition, as an alternative to the input / output device, a serial interface or an Ethernet interface is used as the input / output device, and a display computer having a display or a keyboard or a pointer device is connected to the interface to transmit display information to the display computer. By receiving the input information from the display computer, the display may be performed by the display computer, or by accepting the input, the input and display on the input / output device may be substituted. Further, the display device 400 may be formed integrally with the management computer 100 and the server 200.

The above is the hardware configuration of the computer system according to the first embodiment. A plurality of management computers 100, servers 200, and storage devices 300 may be connected in order to speed up processing and improve reliability.

FIG. 3 shows the hardware configuration of the management computer according to the first embodiment.

The management computer 100 includes a storage resource 110, a communication interface (I / F) 130, and a processor 120 connected to them.

The storage resource 110 is, for example, a memory, and may include an auxiliary storage device such as an HDD. The storage resource 110 stores various information and a program executed by the processor 120. Specifically, as information, VM management table 1101, host management table 1102, storage volume management table 1103, threshold management table 1104, performance management table 1105, countermeasure means management table 1106, metric management table 1107, countermeasure means application. The management table 1108 and the selection means holding table 1109 are stored, and as programs, the monitoring program 1121, the countermeasure plan generation program 1122, the countermeasure policy determination program 1123, the countermeasure means selection program 1124, and the parameter calculation script group 113 are stored.

(1-3) Logical configuration of computer system

FIG. 4 shows the logical configuration of the computer system according to the first embodiment.

The storage device 300 has a plurality of volumes 370. The volume may be, for example, a logical volume, a virtual volume, or a dynamic virtual volume described later.

A logical volume is a volume composed of a physical storage area (called a physical page) of a storage device. Inside the storage device 300, a group of storage devices in which a plurality of storage devices are grouped by RAID (Redundant Arrays of Independent Disks) technology is called a RAID group (not shown). The logical volume is an LU (Logical Unit) that is obtained by cutting out a part of the physical area of the RAID group and making it logically accessible as one storage resource from a computer.

The virtual volume is a virtual volume realized by the thin provisioning mechanism of the storage device 300. The thin provisioning mechanism constitutes a pool (not shown) from logical storage areas (called logical pages) of a plurality of logical volumes. Then, when the thin provisioning mechanism receives the I / O from the server 200 for the virtual volume, the thin provisioning mechanism allocates the logical page from the pool to the virtual volume. As a result, the thin provisioning mechanism can expand the capacity of the virtual volume as needed.

The dynamic virtual volume dynamically changes the physical page corresponding to the logical page to the physical page of another storage device having different I / O performance and reliability in response to a change in the amount of I / O with respect to the virtual volume. It is a virtual volume that can be.

The server 200 is an example of a computer that accesses the storage device 300. The server 200 may have an OS capable of executing an application program or a hypervisor 800 capable of logically generating and executing a VM 700. The hypervisor 800 can control a plurality of VM800s at one time. Each VM800 can execute application programs as if it were a stand-alone physical computer.

The volume 370 is provided for any one of the server 200 and the VM700. For example, the volume 370 provided to the server 200 having a general-purpose OS is recognized by the OS as one storage resource. The volume 370 provided to the server 200 having the hypervisor 800 is recognized as a data store and stores data of a plurality of VM700s. The hypervisor 800 provides the VM 700 by executing the VM data stored in the data store on the processor 210. Therefore, when the setting is changed on the VM700, I / O is generated for the volume 370. Also, the volume 370 may be provided directly to the VM700. In that case, the volume 370 is recognized as one storage resource from the VM700.

The hypervisor 800 may configure the VM700 by using a plurality of volumes 370. In this case, the control information of the VM700, the OS image, the snapshot, and the like are stored in separate volumes 370.

The hypervisor 800 may have a VM migration function of moving a VM 700 running on one server 200 to another server 200. It may also have a VM storage migration function that moves data from one VM 700 to another volume 370.

(1-4) Contents of various tables

FIG. 5 shows the VM management table 1101.

The VM management table 1101 stores various information of the VM 700. From the VM management table 1101, it is possible to know the data store that stores the data of the server 200 that operates the VM 700 and the VM 700, various setting information of the VM 700, the information of the server and the data store that can be selected as the destination of the VM 700, and the like. The management computer 100 periodically communicates with the server 200, acquires various information of each VM 700 operating on the server 200, and stores the information in the VM management table 1101.

Specifically, the VM management table 1101 has the following information in the entry for each VM. The VMID11010 is an ID for identifying the VM. The server ID 11011 is an ID for identifying the server 200 on which the VM is running. The data store name 11012 is the name of the data store in which the data of the VM is stored. The number of CPU cores 11013 is the number of CPU cores assigned to the VM. The memory amount 11014 is the maximum value of the memory amount that can be used by the VM. The number of CPU cores 11013 and the amount of memory 11014 are examples of the set values of the VM700. The VM management table 1101 may include other set values.

The movable server ID 11015 is the ID of the server 200 that can be the destination of the VM. If there is no server 200 that can be the destination of the VM, "-" is stored.

The movable data store name 11016 is the name of a data store that can be the destination of the data of the VM. If there is no data store that can be the destination of the VM data, "-" is stored.

FIG. 6 shows the host management table 1102.

The host management table 1102 stores various information of the computer that accesses the storage device 300. From the host management table 1102, it is possible to determine whether each computer is a physical computer or a VM, and whether a general-purpose OS is running or a hypervisor is running. In addition, various configuration information and setting information of the server 200 can be known. The management computer 100 periodically communicates with the server 200 and the VM 700, acquires various information, and stores it in the host management table 1102.

Specifically, the host management table 1102 has the following information in the entry for each computer. The server ID 11020 is an ID for identifying the computer. The virtualization flag 11021 stores "No" when the computer is the server 200, and stores "Yes" when the computer is the VM700. The VMID11022 stores the VMID11010 when the computer is a VM. If the computer is not a VM, "-" is stored. The OS type 11023 stores the type of OS applied to the computer. When the hypervisor is running on the computer, the "hypervisor" is stored as a value. When a general-purpose OS is running on the computer, a value that can determine the type of OS is stored. For example, the OS name is stored.

The number of CPU cores 11024 stores the number of cores of the processor 210 when the computer is the server 200, and stores the value of the number of assigned cores 11013 when the computer is the VM700. When the computer is the server 200, the memory capacity of the storage resources 220 is stored in the memory amount 11025, and when the computer is the VM700, the value of the allocated memory amount 11014 is stored. The number of CPU cores 11024 and the amount of memory 11025 are examples of computer setting values. The host management table 1102 may include other information such as clustering configuration information and other setting values such as power management settings. The volume ID 11026 stores the identification ID of the volume provided to the computer.

FIG. 7 shows the storage volume management table 1103.

The storage volume management table 1103 stores various information of the volume 370 provided by the storage device 300. From the storage volume management table 1103, the configuration information and the setting information of the volume 370 inside the storage device 300 can be known. For example, you can see the relationship between volume 370 and RAID groups and pools. In addition, the CPU allocation setting for controlling the I / O issued to the volume 370 can be known. Further, the storage volume management table 1103 indicates the computer to which the volume 370 is provided. The management computer 100 periodically communicates with the server 200 and the storage device 300, acquires various information of the volume 370, and stores it in the storage volume management table 1103.

Specifically, the storage volume management table 1103 has the following information in the entry for each volume. The storage ID 11030 is an ID for identifying the storage device 300 having the volume. The volume ID 11031 is an ID for identifying the volume. The RAID group ID 11032 is an ID for identifying the RAID group to which the volume belongs when the volume is a logical volume. If the volume is not a logical volume, the value of "-" is stored. The pool ID 11033 is an ID for identifying the pool to which the volume belongs when the volume is a virtual volume. If the volume is not a virtual volume, the value of "-" is stored.

The responsible CPU ID 11034 is an ID for identifying the processor 310 that is in charge of I / O processing to the volume and other storage internal processing such as replication related to the volume. The cache memory ID 11035 is an ID for identifying the cache memory 380 that is in charge of I / O processing to the volume and other data processing such as replication related to the volume. The cache memory capacity 11036 is the maximum capacity that can be used as a cache in I / O processing to the volume and other data processing such as replication related to the volume. For example, the storage device 300 may have a function of logically dividing the area of the cache memory 380 and allocating an available area for each volume 370. When the area of the cache memory 380 is logically divided, the cache memory capacity 11036 stores the capacity of the cache memory area that can be used by the volume, and the allocation state of the volume to the cache memory area is stored. The value of the cache memory capacity 11036 is changed when is changed. The responsible CPU ID 11034, the cache memory ID 11035, and the cache memory capacity 11036 are a part of the setting information of the volume. The storage volume management table 1103 may include other information, such as settings related to replication of volume 370, and in the case of a dynamic virtual volume, settings such as a storage device on which logical pages are arranged.

The allocation server ID 11037 is an ID that identifies the server 200 or VM700 that receives the provision of the volume. The data store name 11038 is the name of the data store set for the volume on the hypervisor when the volume is provided to the server 200 having the hypervisor.

FIG. 8 shows the threshold management table 1104.

The threshold value management table 1104 holds a threshold value for determining that a performance deterioration to be resolved has occurred for each component constituting the computer system. The threshold value management table 1104 holds two types of threshold values, a warning threshold value and an absolute threshold value. The absolute threshold is a threshold that needs to be maintained in order to protect the SLA (Service Level Agreement). The warning threshold is a threshold for preventing a specific performance metric of a specific resource from approaching the absolute threshold, and is set to a value lower than the absolute threshold. The performance metric is a value indicating the load of the resource and is recorded by the computer system.

Specifically, the threshold value management table 1104 has the following values in the entries for each resource. The resource ID 11040 is an ID that identifies a resource that is a component that constitutes a computer system. The computer system may include internal physical components such as a server 200, a storage device 300, and a physical component constituting the computer system such as a network switch, for example, a RAID group inside the storage device 300 and the like. The computer system may also include logical components with performance metrics, such as VM700. The resource type 11041 indicates the type of the resource. For example, in the case of the pool inside the storage device 300, the value of "Store.Pool" is stored. The performance metric type 11042 represents the type of performance metric for which the threshold value is set. For example, the value of "operating rate" indicating the ratio of the time used per unit time and the value of "IOPS" indicating the number of I / Os per unit time are stored. The warning threshold value 11043 stores the value of the warning threshold value. The warning threshold 11043 stores a value equal to or less than the absolute threshold 11044. The absolute threshold value 11044 stores a threshold value calculated from an SLA or SLO (Service Level Agreement). The warning threshold excess permissible time 11045 stores the time permissible to exceed the warning threshold 11043.

FIG. 9 shows the performance management table 1105.

The performance management table 1105 holds performance data which is time series data of performance metrics of each resource acquired from the computer system. The example of the performance management table 1105 here holds performance data related to the pool inside the storage device 300. The management computer 100 may store a performance management table having a similar structure for each resource type. This figure is shown as an example of them. The performance management table 1105 shows the performance data of each resource at a specific time in the past. The management computer 100 periodically executes the monitoring program 1121 and saves the acquired performance data in the performance management table 1105.

Specifically, the performance management table 1105 has the following information in the entries for each resource and each time. The resource ID 11050 is an ID for identifying the resource. The time 11051 is the time when one value in the performance data of the resource is acquired. The operating rate 11052 is the operating rate of the resource (pool in FIG. 9) at the time. IOPS11053 is the IOPS of the resource (pool in FIG. 9) at the time. Occupancy rates 11052 and IOPS 11053 are examples of performance metrics. These values may be replaced with the performance metrics that can be acquired depending on the resource type, and the types of performance metrics included in the performance management table 1105 are not limited to two.

FIG. 10 shows a countermeasure means management table 1106.

Countermeasure measures The management table 1106 holds countermeasures that can improve the performance metric for each type of bottleneck resource and the type of performance metric that violates the threshold value. The countermeasure measures management table 1106 can be used to determine effective countermeasure measures when performance degradation occurs, and further, it is possible to determine the types of parameter resources that are the parameters of the countermeasure measures and the countermeasure measures that can be combined. ..

Specifically, the countermeasure means management table 1106 has the following information in the entry for each countermeasure means. The bottleneck resource type 11060 is the type of the bottleneck resource that is the target of the countermeasure. A value similar to the resource type 11041 of the threshold management table 1104 can be stored. The bottleneck performance metric type 11061 is a type of performance metric that violates the threshold value. A value similar to the performance metric type 11042 of the threshold management table 1104 can be stored. Countermeasure means 11062 is the countermeasure means capable of improving the performance data of the bottleneck performance metric type 11061 with respect to the bottleneck resource type 11060. For example, a value that can specify a countermeasure means such as "VM storage migration" or "I / O amount upper limit control" is stored. The parameter resource type 11063 is a resource type used as a parameter by the countermeasure means. For example, if the value is "VM", it means that it is a countermeasure measure for VM, and if the value is "Storage.Volume", it means that it is a countermeasure measure for volume 370 inside the storage device 300. Shown. Combineable countermeasure means 11064 indicates another countermeasure that can be implemented at the same time as the countermeasure.

FIG. 11 shows the metric management table 1107.

The metric management table 1107 stores the performance metric type of another resource whose value changes in relation to the performance metric of the bottleneck resource.

Specifically, the metric management table 1107 has the following information in the entry for each combination of performance metrics. Bottleneck resource type 11070 indicates the type of bottleneck resource that is the cause of performance degradation. Bottleneck performance metric type 11071 indicates the type of performance metric of the bottleneck resource that violates the threshold value. The parameter resource type 11072 is a resource type related to the bottleneck performance metric type 11071 of the bottleneck resource type 11070. The parameter performance metric type 11073 is a type of performance metric associated with the bottleneck performance metric type 11071.

An example shown in this figure shows that the operating rate of the pool inside the storage device 300 is increasing (deteriorating) as a result of the increase in IOPS of the VM700 and the volume 370.

FIG. 12 shows a countermeasure means application management table 1108.

The countermeasure measure application management table 1108 can be used to determine whether the countermeasure means are effective or invalid based on the load tendency of the bottleneck resource and the related resource.

Specifically, the countermeasure means application management table 1108 has the following information in the entry for each countermeasure means. The bottleneck resource type 11080 indicates the type of bottleneck resource that causes the performance deterioration. The performance metric type 11081 indicates the type of the performance metric of the bottleneck resource that violates the threshold value. The countermeasure means 11082 stores a value representing a countermeasure means such as VM storage migration.

Countermeasure policy 11083 indicates a policy for determining parameters of the countermeasure means for eliminating the performance deterioration of the bottleneck resource. For example, a policy to reduce the value of performance data at each time point within the specified period on average (described as "average reduction" in this figure), or the value of performance data at the time of peak value within the specified period. The policy of preferentially lowering (described as "peak lowering" in this figure) is stored. In addition, a policy other than the above two policies may be described.

Related resource tendency 11084 shows the load tendency of the related resource related to the bottleneck resource in the specified period. The load tendency is represented by one of a plurality of patterns. For example, the fluctuation range of performance data between multiple time points in a specified period is small (described as "stable" in this figure), or only the value of performance data at a specific time point is large ("not" in this figure). A value indicating a pattern such as "stable") is stored. In addition. A value representing a tendency other than the above two patterns may be stored.

Applicability 11085 indicates whether it is generally effective or invalid to apply the countermeasure means 11082 in the countermeasure policy 11083 when the load tendency of the related resource is the pattern of the related resource tendency 11084 with respect to the resource type of the bottleneck resource type 11080. Indicates. If it is valid, "Yes" is stored, and if it is invalid, "No" is stored. When the applicability 11085 is "No", it is indicated that the effect of performance improvement is unlikely to be obtained even if the countermeasure measures are implemented in consideration of the load tendency of the bottleneck resource and the related resource.

The priority 11086 indicates the priority of the countermeasure plan using each countermeasure means when the load tendency of the related resource is the pattern of the related resource tendency 11084 with respect to the resource type of the bottleneck resource type 11080. For example, it indicates that a countermeasure plan having a priority of 1 should be implemented with priority over a countermeasure plan having a priority of 2. The priority 11086 may be determined, for example, by the degree of influence on the performance of each resource, the degree of influence on reliability, and the like.

The script ID 11087 is an ID that identifies a script that describes a method for calculating the parameters of the countermeasure means 11082. A specific example of the script will be described later.

FIG. 13 shows the selection means holding table 1109.

The selection means holding table 1109 holds the countermeasure means determined by the countermeasure means selection program 1124 and the parameter candidate resource which is a resource that is a parameter candidate of the countermeasure means. Parameter candidate resources are at least part of the associated resource for the bottleneck resource.

Specifically, the selection means holding table 1109 has the following information in the entry for each plan. The plan shows a set of countermeasures and parameter candidate resources. Plan ID 11090 is an ID that identifies a plan for a specified period, bottleneck resource, and performance metric. When multiple countermeasures can be taken for a specified period, bottleneck resource, and performance metric, priority is given to the application of each countermeasure when combining the parameter candidate resources of each countermeasure and the countermeasures for one plan. The degree is stored.

Bottleneck resource type 11091 indicates the type of bottleneck resource that is the cause of performance degradation. The bottleneck resource ID 11092 is an identification ID of the bottleneck resource that causes the performance deterioration. The performance metric type 11093 indicates the type of the performance metric of the bottleneck resource that violates the threshold value. The designated period 11094 is a designated period in which the generation of a countermeasure plan is designated. Countermeasure policy 11095 indicates a policy for determining parameters of countermeasure means for eliminating the performance deterioration of bottleneck resources. The countermeasure policy 11095 stores the same value as the countermeasure policy 11083. Related resource trend 11096 shows a pattern of resource load tendency related to bottleneck resources. The related resource tendency 11096 stores the same value as the related resource tendency 11084. The countermeasure means 11097 stores a value representing the countermeasure means such as VM storage migration. The countermeasure means 11097 stores the same value as the countermeasure means 11082. When a plurality of countermeasure measures 11097 corresponding to the same plan ID 11090 are combined to generate a countermeasure plan, priority 11098 indicates an application order of each countermeasure means. The parameter candidate resource 11099 stores an ID that identifies the parameter candidate resource of the countermeasure means 11097. In the case of "All", it means that all resources that can be taken as parameters of the countermeasure means 11097 are candidates.

(1-5) Details of operation of each device

Next, the performance monitoring process of the management computer 100 will be described. The performance monitoring process is a process executed by executing the monitoring program 1121 stored in the management computer 100 on the processor 120.

FIG. 14 is a flowchart of the performance monitoring process.

The performance monitoring process may be executed periodically or may be executed according to an instruction from the user. The performance monitoring process of this embodiment is periodically executed at preset monitoring intervals. The performance monitoring process acquires performance data from various components constituting the computer system, and determines whether or not it is necessary to implement performance measures in light of the threshold information registered in the threshold management table 1104. If performance countermeasures are required, the performance monitoring process executes the countermeasure plan generation program 1122.

In this embodiment, the cases where it is judged that the performance countermeasure is necessary are the case where a certain time has passed since the warning threshold was exceeded and the absolute threshold when the countermeasure plan is generated after the next performance monitoring process is executed. Two cases, one with a high possibility of exceeding the threshold, and the other with a high possibility of exceeding the threshold value will be described.

First, the monitoring program 1121 refers to a table that manages configuration information of resources constituting the computer system, such as the VM management table 1101, the host management table 1102, and the storage volume management table 1103, and for each resource under management. The performance data of the current time is acquired and stored in the performance management table 1105 (S1000, S1010). The monitoring program 1121 performs the processes of S1000 to S1110 for each resource.

Next, the monitoring program 1121 refers to the threshold value management table 1104, and warns the warning threshold value 11043 and the absolute threshold value 11044 for each performance metric type 1142 based on the resource type and the resource ID. Acquire the threshold value excess allowable time 11045 (S1020). If the warning threshold 11043 is not set, the monitoring program 1121 does not need to perform subsequent processing for performance violation check, so it returns to S1000 and performs processing for the next resource (S1030). ..

If the warning threshold 11043 is set, the monitoring program 1121 compares the performance data of the current time acquired in S1010 with the warning threshold 11043, and if the warning threshold 11043 is not violated, Return to S1000 and perform processing for the next resource (S1040).

As a result of comparison in S1040, if the warning threshold value 11043 is violated, whether or not the monitoring program 1121 is allowed to exceed the warning threshold value 11043 until the next performance monitoring process is executed. To judge. Therefore, the monitoring program 1121 first acquires the warning threshold excess allowable time 11045 from the threshold value management table 1104 (S1050). Next, the monitoring program 1121 estimates the estimated generation time of the countermeasure plan (S1060). As a method, for example, for the resource, the past countermeasure generation time is stored for the countermeasures generated based on the same type and the same number of related resources, and the countermeasure proposal generation time is used as the expected countermeasure generation time. There is.

Next, the monitoring program 1121 calculates whether the warning threshold excess permissible time 11045 is larger or smaller than the value obtained by adding the estimated generation time of the countermeasure proposal estimated in S1060 and the execution interval of the monitoring program 1121. (S1070). If the warning threshold exceeding time 11045 is smaller (when the result of S1070 is Yes), the monitoring program 1121 will always exceed the warning threshold until the next monitoring program 1121 is executed. This means that the excess allowable time 11045 will be exceeded. Therefore, the monitoring program 1121 determines that a performance abnormality has occurred in the performance metric of the resource (S1080).

If the result of S1070 is No, then the monitoring program 1121 determines whether or not the absolute threshold 11044 is set for the resource (S1090). When the absolute threshold is set (Yes), the monitoring program 1121 estimates the future performance change based on the past load tendency of the resource and calculates the estimated time to reach the absolute threshold (S1100). ). As a method of predicting future performance changes, for example, an approximate expression of performance change is calculated using the least squares method for performance data over a certain period in the past, and it is assumed that the past trend will continue in the future, and it is absolute from the approximate expression. A method of calculating the time when the threshold value 11044 is reached may be used.

Next, in the monitoring program 1121, whether the absolute threshold arrival estimated time estimated in S1100 is larger or smaller than the value obtained by adding the countermeasure proposal generation estimated time estimated in S1060 and the execution interval of the monitoring program 1121. Is calculated (S1100). If the estimated time to reach the absolute threshold is smaller (when the result of S1110 is No), the monitoring program 1121 is likely to exceed the absolute threshold 11044 by the time the next monitoring program 1121 is executed. Means. Therefore, the monitoring program 1121 determines that a performance abnormality has occurred in the performance metric of the resource (S1080).

When the result of the determination in S1090 is No and the result of the determination in S1110 is Yes, the monitoring program 1121 determines that no performance abnormality has occurred in this embodiment. In this embodiment, there is a high possibility that the absolute threshold value will be exceeded when a certain time has passed since the warning threshold value was exceeded and when a countermeasure plan is generated after the next performance monitoring process is executed. It was judged that a performance abnormality occurred when the two cases corresponded to the case. However, the monitoring program 1121 may determine that a performance abnormality has occurred in other cases, for example, in cases where the threshold value has been violated more than a certain number of times in the past. When considering cases other than the two cases mentioned in this embodiment, the monitoring program 1121 further determines the performance abnormality when the judgment result of S1090 is No and when the judgment result of S1110 is Yes. Processing may be added and executed.

As a result of performing the processes from S1000 to S1110 for all resources, if no performance abnormality has occurred, it is not necessary to generate a countermeasure plan, so the monitoring program 1121 ends (S1120).

When a performance abnormality has occurred for one or more resources (when the result of S1120 is Yes), the monitoring program 1121 performs the subsequent processing and executes the countermeasure plan generation program 1122.

Since each resource that composes a computer system is related to each other, there may be a resource whose performance is secondarily deteriorated due to the performance deterioration of a specific resource. Therefore, the monitoring program 1121 determines in S1130 whether the performance deterioration of each resource is caused by the performance change of another resource, and extracts the bottleneck resource and the performance metric which are the final causes. As a method of this cause analysis, for example, a public technique of RCA (Root Cause Analysis) may be used.

Next, the monitoring program 1121 sets the earliest time that exceeds the warning threshold among the resources determined to be performance abnormal as the start time and the current time as the end time for the performance metric of the bottleneck resource extracted in S1130. The countermeasure plan generation program 1122 is executed by designating the period from the start time to the end time as the designated period, and the countermeasure plan for the designated period is generated. Then, an alert is displayed on the display device 400 together with the generated countermeasure plan.

The longer the monitoring interval and the specified period, the better the accuracy of the countermeasure plan, but the more likely it is that the threshold will be exceeded. According to this performance monitoring process, it is possible to prevent the performance metric of the bottleneck resource from exceeding the threshold value and improve the accuracy of the countermeasure plan.

According to this performance monitoring process, the management computer 100 periodically monitors the load of the resource so that the time when the load exceeds the warning threshold does not exceed the warning threshold exceeding allowable time and the load exceeds. A specified period can be set so that the absolute threshold is not exceeded, and a countermeasure plan for the specified period can be generated.

Next, the countermeasure plan generation process of the management computer 100 will be described. The countermeasure plan generation process is a process executed by executing the countermeasure plan generation program 1122 stored in the management computer 100 on the processor 120.

FIG. 15 is a flowchart of the countermeasure plan generation process.

The countermeasure plan generation process may be executed according to an instruction from the user, or may be executed by the countermeasure plan generation program 1122 when a performance abnormality is detected.

The countermeasure plan generation program 1122 first receives the bottleneck resource type, the bottleneck resource identification ID, the designated period for generating the countermeasure plan, and the type of the performance metric that violates the warning threshold 11043 (S2000). The above information may be specified by the user from the display device 400, or may be specified by the monitoring program 1121. It should be noted that these are examples in this embodiment, and do not limit that the countermeasure plan generation program 1122 is called by another program.

Next, the countermeasure plan generation program 1122 implements the countermeasure policy determination process (S2020). The countermeasure policy judgment process is a process of analyzing the load tendency of bottleneck resources in a specified period and requesting a countermeasure policy for efficiently generating an accurate countermeasure plan for the problem of performance deterioration. A specific flowchart will be described later.

Next, the countermeasure plan generation program 1122 executes the countermeasure means selection process (S2030). The countermeasure selection process is a process of analyzing the load tendency of the related resource group related to the bottleneck resource in a specified period and selecting an appropriate countermeasure for the response policy. A specific flowchart will be described later.

The countermeasure plan generation program 1122 acquires the result of the countermeasure means selection process from the selection means holding table 1109 (S2040). At this time, since there may be a plurality of countermeasures for eliminating the performance deterioration of the bottleneck resource, the bottleneck resource, the performance metric, and the designation for which a plurality of countermeasures (plans) are specified in the selection means holding table 1109. There is a possibility that multiple plans corresponding to the period are held. Therefore, the countermeasure plan generation program 1122 executes the subsequent processes (S2070 to S2110) for each plan.

The plan 11090 may be a combination of a plurality of countermeasures. In that case, the countermeasure plan generation program 1122 refers to the priority 11098 for each countermeasure means 11097 constituting the plan 11090, and executes the subsequent processes (S2070 to S2110) in the order of the countermeasure means 11097 having the highest priority.

First, the countermeasure plan generation program 1122 selects the countermeasure means 11097 from the selection means holding table 1109 in S2070, and acquires the corresponding countermeasure policy 11095, the related resource tendency 11096, and the parameter candidate resource 11099.

Next, in S2080, the countermeasure plan generation program 1122 refers to the countermeasure means application management table 1108 based on the information acquired in S2070, and acquires the script ID 11087 that describes the method of calculating the parameters of the countermeasure means.

In S2090, the countermeasure plan generation program 1122 specifies the script corresponding to the script ID 1087 acquired in S2080 from the parameter calculation script group 113 stored in the storage resource 110. Then, the countermeasure plan generation program 1122 calculates the parameter using the parameter candidate resource acquired in S2070 and the specified script.

The countermeasure plan generation program 1122 estimates the change in the value of the performance metric that caused the threshold violation of the bottleneck resource when the countermeasure plan calculated by the script is implemented (S2100). This estimation method may be the same as the estimation method of S1100, or may be estimated by applying a countermeasure plan to the result estimated by S1100. If, as a result of the estimation, the warning threshold value 11043 is not violated at all time points within the specified period (the result of S2110 is Yes), it is not necessary to implement countermeasures having a lower priority of 11098. The countermeasure plan generation program 1122 starts the countermeasure plan generation process (S2070 to S2110) for the next plan 11090. If the result of S2110 is No, the countermeasure plan generation program 1122 executes the countermeasure plan generation process (S2070 to S2110) for the countermeasure means having a lower priority of 11098.

After that, the countermeasure plan generation program 1122 displays the generated countermeasure means on the display device 400 for each plan (S2120). As a display method, it may be expressed in a standard format such as XML. For example, in that case, not only the display device 400 may be displayed, but also the file may be electronically stored in a format such as a file.

Further, the management computer 100 may represent the estimate when the generated countermeasure plan is implemented as a graph, and have the display device 400 present it together with the countermeasure plan. This makes it possible to intuitively grasp the changes caused by the implementation of the countermeasure plan.

Further, the generated countermeasure plan may be expressed in the form of CLI corresponding to the countermeasure means. Along with this, in this embodiment, the display of contents other than the countermeasure means and the parameters is not limited. For example, the countermeasure plan may be expressed in the form of a program as long as the content of the countermeasure plan is included.

According to this countermeasure plan generation process, the management computer 100 can determine the parameters by using an appropriate script, and can generate the countermeasure plan for the designated period.

The scripts in this embodiment are Script1 to Script4. Script1 to Script4 are scripts for generating a countermeasure plan to reduce the operating rate of the pool to the warning threshold 11043 when the operating rate of the pool in the storage device 300 has increased as the cause of the performance problem. is there. The parameter candidate resource here is a VM. In particular, Script1 and Script2 are scripts that describe a calculation method for calculating the parameters of VM storage migration. Script3 and Script4 are scripts that describe a calculation method for calculating parameters for a storage setting change that limits the upper limit of the I / O amount of a volume.

FIG. 16 is a flowchart of Script1.

Script1 determines the VM700 that performs VM storage migration and the data store of the migration destination. Specifically, Script1 is a combination of VM700 that can reduce the average utilization rate of bottleneck resources in a specified period to a value smaller than the warning threshold 11043, when the data of VM700 is moved. Find the combination that minimizes the average utilization rate of each pool or RAID group during the specified period. This may be modeled as, for example, an integer programming problem and calculated using a known integer programming problem solver or the like. The reason why this is possible is that the optimization problem can be converted into a simpler problem while maintaining the effect of eliminating the performance deterioration by the countermeasure policy judgment process. If the number of resources is still large and the optimum solution cannot be generated in a realistic time, for example, the above-mentioned integer programming problem may be linearly relaxed and an approximate solution may be obtained using a known algorithm.

When the utilization rate of the pool exceeds the threshold value over a specified period due to the total load of a plurality of VMs, the VMs can be moved so that the load is not concentrated on a specific resource by using Script1.

FIG. 17 is a flowchart of Script2.

Script2 determines the VM700 that performs VM storage migration and the data store of the migration destination. Specifically, in Script2, among the combinations of VM700 that can reduce the maximum utilization rate of the bottleneck resource in the specified period to a value smaller than the warning threshold 11043, each of the combinations of VM700 after moving the data of VM700. Find the combination that minimizes the maximum utilization rate of the pool or RAID group in the specified period. This may be modeled as, for example, an integer programming problem and calculated using a known integer programming problem solver or the like. The reason why this is possible is that the optimization problem can be converted into a simpler problem while maintaining the effect of eliminating the performance deterioration by the countermeasure policy judgment process. If the number of resources is still large and the optimum solution cannot be generated in a realistic time, for example, the above-mentioned integer programming problem may be linearly relaxed and an approximate solution may be obtained using a known algorithm.

According to Script2, when the utilization rate of the pool exceeds the threshold value at a specific time during the specified period, the VM can be moved so as to minimize the peak value of the destination resource.

FIG. 18 is a flowchart of Script3.

Script3 determines the upper limit of I / O for the volume 370 that is loading the bottleneck resource. The Script3 first calculates the upper limit of the IOPS value of the pool for the bottleneck resource (pool in the storage device 300) so that the operating rate becomes the warning threshold value 11043 or less (S11330).

Actually, the target for issuing I / O by the server 200 or VM700 is the volume 370, and the amount of I / O issued to the pool is the I / O issued to the volume 370 generated from the pool. It is the total amount of / O. Therefore, Script3 determines the IOPS upper limit of each volume 370 so that the total IOPS of each volume 370 is equal to the IOPS upper limit of the pool calculated in S11330. At that time, for example, the average IOPS of each volume 370 in the designated period is calculated, and the IOPS upper limit of the pool is distributed according to the ratio to calculate the IOPS upper limit of each volume 370.

If the volume load is stable over the specified period and the pool utilization exceeds the threshold over the specified period, according to Script3, each will reduce the volume load over the specified period. The IOPS upper limit of the volume 370 can be determined.

FIG. 19 is a flowchart of Script4.

Script4 determines the upper limit of I / O for the volume 370 that is loading the bottleneck resource. In Script4, the upper limit of the IOPS value of the pool, which is a bottleneck resource, is calculated in the same process as S11330 (S11340).

Next, the countermeasure plan generation program 1122 sets the IOPS value calculated in S11340 as the upper limit value for the volume 370 related to the pool which is a bottleneck resource (S11341).

When the capacity of the pool exceeds the threshold value due to the load of the volume becoming high at a specific point in the specified period, according to Script4, the IOPS upper limit is determined so as to suppress the load of the specific volume. be able to.

The above script is an example, and scripts other than the above may be included depending on the bottleneck resource, the performance metric, the countermeasure means, and the countermeasure policy. Further, although it is expressed in the form of a flowchart in this embodiment, it may be expressed in another form in which the processing contents and the processing order can be expressed, for example, in the form of XML (Extension Markup Language). Further, the management computer 100 may have a function of dynamically reading XML data contained in an electronic medium such as a file into a storage resource.

FIG. 20 is a flowchart showing a countermeasure policy determination process (S2020).

The countermeasure policy determination process is executed by executing the countermeasure policy determination program 1123 on the processor 120.

First, the countermeasure policy determination program 1123 receives the type of the bottleneck resource, the ID of the bottleneck resource, the specified period, and the type of the performance metric that violates the warning threshold value 11043 (S3000).

Next, the countermeasure policy determination program 1123 refers to the performance management table 1105 corresponding to the bottleneck resource, and obtains the specified bottleneck resource ID, the performance metric type, and the performance data at each time point included in the specified period. Acquire (S3010).

Next, the countermeasure policy determination program 1123 classifies the tendency of the performance data of the bottleneck resource acquired in S3010 into a plurality of patterns. In this embodiment, as an example of classification, the case where the value of the performance data at each time point has little variation is regarded as a "stable" pattern, and conversely, the value of the performance data at a specific time point is the value of the performance data at another time point. The case of classifying into two patterns will be described, in which the case where the pattern is significantly different from the above is regarded as the “unstable” pattern.

The countermeasure policy determination program 1123 calculates the average value of the performance data and the bottleneck load deviation threshold value over the specified period in order to determine whether the performance data acquired in S3010 is a stable pattern or an unstable pattern. For example, the countermeasure policy determination program 1123 calculates the standard deviation of the performance data over a specified period, and calculates the bottleneck load deviation threshold value by multiplying the standard deviation by a preset coefficient. Then, the countermeasure policy determination program 1123 determines whether or not there is a time point in which the deviation of the performance data value with respect to the average value becomes equal to or greater than the bottleneck load deviation threshold value with respect to the performance data value at each time point (S3020, S3030). If the result of S3030 is Yes, the countermeasure policy determination program 1123 determines that the pattern is unstable, and if No, it determines that the pattern is stable. The calculation method using the standard deviation used in S3030 is an example of a method for obtaining outliers, and other outlier detection methods may be used.

Then, when the pattern is classified as unstable as a result of S3030, it is highly probable that a performance abnormality has occurred due to a large threshold violation at a specific time point in the specified period. Therefore, the countermeasure policy determination program 1123 determines to adopt a countermeasure policy for preferentially improving the value of the performance data at the time when the peak value is taken in the designated period (S3040). In the following explanation, this policy will be described as "peak reduction".

On the other hand, if it is classified as a stable pattern as a result of S3040, it is highly likely that the threshold value is constantly violated during the designated period. Therefore, the countermeasure policy determination program 1123 is used during the designated period. It is decided to take a countermeasure policy to uniformly improve the value of the performance data at each time point (S3050). In the following explanation, this policy will be described as "average reduction".

According to this countermeasure policy determination process, it is possible to determine whether the load tendency of the bottleneck resource is stable or unstable, and determine an appropriate countermeasure policy based on the determination result.

FIG. 21 is a flowchart showing a countermeasure means selection process (S2030).

The countermeasure means selection process is executed by executing the countermeasure means selection program 1124 on the processor 120.

First, the countermeasure means selection program 1124 receives the bottleneck resource type, the bottleneck resource ID, the specified period, the performance metric type that violates the warning threshold 11043, and the countermeasure policy ( S4000).

Next, the countermeasure measure selection program 1124 extracts countermeasure measures effective in improving the performance data for the performance metric of the specified bottleneck resource type (S4010). Specifically, the countermeasure means selection program 1124 refers to the countermeasure means management table 1106, and corresponds to the bottleneck resource type 11060 received in S4000 and the bottleneck performance metric type 11061 causing the warning threshold violation. The countermeasure means 11062, the parameter resource type 11063, and the countermeasure means 11064 that can be combined are extracted.

Here, there is a possibility that a plurality of countermeasures 11062 can be applied to the bottleneck resource. Further, since the parameter resource type differs depending on each countermeasure means, the countermeasure means selection program 1124 executes the subsequent processing (S4030 to S4120) for each countermeasure means.

The countermeasure means selection program 1124 corresponds to the parameter resource type of the countermeasure means acquired in S4010 from various tables that manage configuration information such as the VM management table 1101, the host management table 1102, and the storage volume management table 1103 in S4030. Extract resource information (S4030). For example, when the bottleneck resource is the pool inside the storage device 300, it is shown that two countermeasure measures 11062, VM storage migration and I / O amount upper limit control, can be taken as countermeasure measures. It is shown that the VM 700 needs to be set as a parameter in order to execute the VM storage migration, and the volume 370 needs to be set as a parameter in order to execute the I / O amount upper limit control.

Here, when the information of the related resource of the corresponding parameter resource type does not exist in the table for managing various configuration information (when the result of S4040 is No), the countermeasure means selection program 1124 may generate a countermeasure plan. Since it cannot be done, the process returns to S4020 and the same processing is performed for the next countermeasure means.

When the result of S4040 is Yes, the countermeasure means selection program 1124 classifies the load tendency of each related resource into patterns by the subsequent processing. First, the countermeasure means selection program 1124 refers to the metric management table 1107 in order to specify the performance metric of interest in the related resource, and corresponds to the bottleneck resource type 11070, the bottleneck performance metric type 11071, and the parameter resource type 11072. The parameter performance metric type 11073 is extracted (S4050). For example, the utilization rate, which is a performance metric of the pool in the storage device 300, indicates that it is related to the IOPS, which is the performance metric of the VM700, or the IOPS, which is the performance metric of the volume 370.

Next, the countermeasure means selection program 1124 analyzes the load tendency for each related resource. In this embodiment, as in the countermeasure policy determination program 1123, a case where each related resource is classified into a stable pattern and an unstable pattern will be described. Specifically, the countermeasure means selection program 1124 processes S4070 and S4080 for each related resource.

The countermeasure means selection program 1124 refers to the performance management table 1105 in S4070, and extracts the performance data of the parameter performance metric type 11073 of the related resource in the specified period. Then, in S4080, by performing outlier determination on the performance data extracted in S4070, it is determined whether or not the related resource has a stable pattern. The process performed here may be the same as in S3020 (S4080). For example, the countermeasure means selection program 1124 calculates the average value of the performance data and the related load deviation threshold value over a specified period, and when the deviation of the performance data value with respect to the average value becomes equal to or larger than the related load deviation threshold value. Determine if there is. The associated load deviation threshold is, for example, a value obtained by multiplying the standard deviation by a preset coefficient.

After classifying all the related resources extracted in S4030 into patterns, the countermeasure selection program 1124 indicates whether the number of related resources classified into stable patterns is larger than the number of related resources classified into unstable patterns. It is determined whether or not (S4090). If the result is Yes, the countermeasure means selection program 1124 determines that the related resource group that is the parameter of the countermeasure means has a stable pattern for the specified bottleneck resource (S4110). On the contrary, if No, the countermeasure means selection program 1124 determines that the related resource group has an unstable pattern (S4100).

Next, the countermeasure measure selection program 1124 collates the countermeasure policy received in S4000 with the pattern of the related resource determined in S4100 or S4110, and is the countermeasure measure effective in eliminating the performance deterioration of the bottleneck resource? Judge whether or not. Specifically, the countermeasure means selection program 1124 refers to the countermeasure means application management table 1108, and sets the bottleneck resource type 11080, the performance metric type 11081, the countermeasure means 11082, the countermeasure policy 11083, and the related resource tendency 11084. Acquires the matching applicability 11085 value. If the result is Yes, the countermeasure means selection program 1124 determines that the countermeasure means is effective, and if No, determines that the countermeasure means is not effective (S4120).

For example, when the countermeasure policy 11083 is lowered on average and the related resource tendency 11084 is an unstable pattern, the load is increasing at different points in time for each related resource, so that when each related resource is assembled, it is stable throughout the specified period. It is considered that the load tendency seems to have increased. Therefore, if the countermeasure 11082 is VM storage migration at this time, even if the data of a specific VM700 is moved to a different pool, a warning threshold violation occurs in the destination pool at the timing when the load of the VM700 increases. Probability is high. Therefore, applicability 11085 is determined to be No.

Further, when the countermeasure policy 11083 has a peak decrease and the related resource tendency 11084 has an unstable pattern, the load of each related resource is increasing at the same time point, and when each related resource is aggregated, a large bottleneck resource is obtained at a specific time point. It is considered that the load on the vehicle is increasing. Therefore, at this time, when the countermeasure means 11082 is the I / O amount upper limit control, it is necessary to similarly limit the I / O amount to each volume 370, which is large for the application using each volume 370 at the peak time. It will cause an impact. Therefore, the applicability is judged as No.

According to this countermeasure selection process, it is possible to determine whether the load tendency of related resources is stable or unstable, and determine an appropriate countermeasure based on the determination result and the countermeasure policy. it can.

As described above, according to the present embodiment, the management computer 100 can determine the generation policy of the countermeasure plan corresponding to the tendency of the load change of the bottleneck resource in the designated period.

Further, according to the present embodiment, the management computer 100 can determine one load tendency pattern expressing the characteristics of the related resource group. As a result, the management computer 100 can exclude the countermeasure measures that increase the calculation amount of the parameter calculation and the countermeasure measures that have a large influence on the application due to the large side effects.

In addition, the management computer 100 calculates the parameters of the countermeasure means based on the countermeasure policy derived from the load tendency of the bottleneck resource, so that the parameters necessary for eliminating the performance deterioration can be calculated with the minimum calculation. It becomes possible to ask.

(2-1) Outline of the embodiment of this embodiment

In this embodiment, by classifying the related resources of the bottleneck resource for each pattern and deriving the countermeasures applicable to the pattern, it is possible to derive a countermeasure plan combining a plurality of countermeasures. As a result, in the second embodiment, the management computer 100 reduces the amount of calculation of parameter calculation even when related resources of a plurality of patterns are mixed, and has a small side effect on the application and each VM. Can be generated.

Since the configuration of the computer system according to the second embodiment and the information managed by the computer system are the same as those of the computer system described in the first embodiment, the illustration is omitted. Further, among the programs executed by the computer system according to the second embodiment, the operations of the programs other than the countermeasure means selection program 1124 are the same as those described in the first embodiment. Therefore, in the following, among the operations of the countermeasure means selection program 1124, the points different from those described in the first embodiment will be mainly described.

(2-2) Details of operation of each device

FIG. 22 is a flowchart showing a countermeasure means selection process (S2030) according to the second embodiment.

First, the countermeasure means selection program 1124 performs the processes up to S4010 of the first embodiment. After that, the countermeasure means selection program 1124 performs the same processing as in S4030 to S4050 of the first embodiment for each countermeasure means. Further, the countermeasure means selection program 1124 then performs the same processing as in S4070 to S4080 of the first embodiment for each related resource.

After that, in the present embodiment, the countermeasure means selection program 1124 determines whether the related resource has a stable pattern or an unstable pattern as a result of the outlier determination of S4080. In this embodiment, it is assumed that the outliers are determined by the same method as in Example 1 (S3020, S4080) (S5000). If the result of S5000 is Yes, the countermeasure means selection program 1124 determines that the related resource has a stable pattern, and stores the related resource in the storage resource as one of the related resource groups of the stable pattern (S5010). ). On the other hand, if the result of S5000 is No, the countermeasure means selection program 1124 determines that the related resource has an unstable pattern, and the countermeasure means selection program 1124 determines that the related resource is an unstable pattern related resource. It is stored in a storage resource as one of the groups (S5020).

When the countermeasure means selection program 1124 executes the processing up to S5020 for each related resource, the resource having the largest number of related resources among the related resource groups corresponding to each pattern classified in S5010 and S5020 next. Select a group (S5030).

Then, the countermeasure means selection program 1124 sets the countermeasure means application management table 1108 to determine whether or not the countermeasure means is effective in eliminating the performance deterioration of the bottleneck resource with respect to the tendency of the related resource group selected in S5030. Judgment is made by reference (S5040). The content of the judgment and specific examples are the same as in S4120.

If the result determined in S5040 is Yes, the countermeasure is a countermeasure that can be applied to the largest number of related resources among the resources related to the problem of performance deterioration of the bottleneck resource. Therefore, when the countermeasures are combined, it is considered that the countermeasures have the greatest effect of eliminating the performance deterioration of the bottleneck resource. Therefore, the countermeasure means selection program 1124 stores the countermeasure means in the storage resource as the countermeasure means having the highest priority (priority 1), and uses it for the countermeasure means combination generation process described later (S5060).

If the result determined in S5050 is No, the countermeasure means is a countermeasure measure that cannot be applied to the largest number of related resources among the related resources of the bottleneck resource. Therefore, the countermeasure means selection program 1124 determines that the problem of performance deterioration cannot be solved even if the countermeasure means is combined with other countermeasure means, and ends the process for the countermeasure means.

Next, the countermeasure means selection program 1124 determines other countermeasure means to be combined with the countermeasure means based on the pattern of the related resource group other than the largest number of related resource groups by the countermeasure combination generation process (S5070). The details of the countermeasure combination generation process will be described later.

Finally, the countermeasure means selected in S5060 and the countermeasure means determined in S5080 are registered in the selection means holding table 1109. At the time of registering the countermeasure means 11097, both the priority selected in S5060 as the corresponding priority 11098 and the priority selected in the countermeasure means combination generation process are set. Further, when the countermeasure means 11097 is registered, the related resource included in the related resource group corresponding to the countermeasure means is set as the corresponding parameter candidate resource 11099.

FIG. 23 is a flowchart showing a countermeasure means combination generation process according to the second embodiment.

The countermeasure means combination generation process is a subprogram called from the countermeasure means selection program 1124.

In the countermeasure combination generation process, first, the bottleneck resource type and ID, the performance metric type causing the warning threshold violation, the countermeasure policy, the selected countermeasure, and the association classified by S5010 or S5020. Receives the resource group (S6000).

After that, in the countermeasure means combination generation process, among the related resource groups received in S6000, the related resource groups excluding the related resource group having the largest number of resources are selected in descending order of the number of resources, and the subsequent processing is performed (S6010). ..

First, in the countermeasure means combination generation process, the countermeasure means management table 1106 is referred to, and the countermeasure means 11064 that can be combined is acquired for the countermeasure means received in S6000 and all the countermeasure means of the countermeasure means selected in S6040. Then, the countermeasure measures commonly included in all the acquired countermeasure measures 11064 that can be combined are extracted (S6020).

Further, the countermeasure means combination generation process is performed in S6020 by using the bottleneck resource type 11080 received in S6000, the performance metric type 11081, the countermeasure policy 11083, and the related resource tendency 11084 of the related resource group selected in S6010. For each of the extracted countermeasure measures, the countermeasure measure application management table 1108 is referred to, and effective countermeasure measures for eliminating the performance deterioration are extracted (S6030).

Next, the countermeasure means combination generation process multiplies each countermeasure measure extracted in S6030 with the combination of the countermeasure means generated by the previous process of S6040 to generate a combination, and newly stores the generated combination. (S6040).

If the processes from S6010 to S6040 are performed for all the related resource groups, the countermeasure means combination generation process ends the process, and the generated combination of countermeasure measures is returned to the caller of the process. If not, the countermeasure means combination generation process shifts the process to S6010, and then performs the processes from S6020 to S6040 for the related resource group having the largest number of resources.

If you try to generate an appropriate countermeasure plan from a large number of configuration change means or a combination of parameters of the configuration change means so as to prevent performance deterioration at all points in the period, it will be displayed in a realistic calculation time. Can't. According to each of the above-described embodiments, it is possible to limit the candidates for the countermeasure means and reduce the calculation amount of the parameter calculation of the countermeasure means. As a result, it is possible to efficiently generate and display a countermeasure plan in which the performance deterioration is not reproduced when the load is reproduced for the past period.

The past period corresponds to the designated period, etc. The bottleneck load data corresponds to the performance data of the bottleneck resource. The related load data corresponds to the performance data of related resources. The bottleneck load range corresponds to the range in which the deviation from the average value of the performance data of the bottleneck resource is equal to or less than the bottleneck load deviation threshold value. The related load range corresponds to a range in which the deviation from the average value of the performance data of the related resource is equal to or less than the related load deviation threshold value. The plurality of countermeasure policy candidates correspond to the values and the like shown in the countermeasure policy 11083. The plurality of countermeasure measure candidates correspond to the values and the like shown in the countermeasure means 11062. The plurality of determination results correspond to the values and the like shown in the related resource tendency 11084.

Although the embodiments of the present invention have been described above, this is an example for explaining the present invention, and does not mean that the scope of the present invention is limited to the above configuration. The present invention can also be implemented in various other forms.

10 ... Computer system, 100 ... Management computer, 110 ... Storage resource, 120 ... Processor, 200 ... Server, 210 ... Processor, 220 ... Storage resource, 300 ... Storage device, 310 ... Processor, 320 ... Memory, 360 ... Storage device group , 370 ... Volume, 380 ... Cache Memory, 390 ... Controller, 400 ... Display Device, 500 ... Communication Network, 600 ... Communication Network, 700 ... VM, 800 ... Hyper Processor

Claims (12)

It is a management method of a computer system using a management computer.
The management computer
The relative causes a is the bottleneck resource degradation of the computer system of the prior period among a plurality of resources in the computer system, the bottleneck load data is time-series data of the load of the bottleneck resource of the period Acquired,
For the related resource that affects the load of the bottleneck resource among the plurality of resources, the related load data which is the time series data of the load of the related resource during the period is acquired.
When the bottleneck load range is calculated based on the bottleneck load data and all the values in the bottleneck load data are within the bottleneck load range, the tendency of the bottleneck load data is stable. Judge,
When it is determined that the tendency of the bottleneck load data is stable, the performance degradation can be resolved from among a plurality of countermeasure policy candidates indicating the policy for determining the parameters necessary for operating the computer system. Select a countermeasure policy and
Based on the countermeasure policy and the tendency of the related load data, the countermeasure means for resolving the performance deterioration is selected from a plurality of countermeasure measure candidates indicating the operation of the computer system.
The parameters of the countermeasure means are determined based on the countermeasure policy .
Management method. Said management computer, if the tendency of the previous SL bottleneck load data is determined to be stable, based on the result of the determination of the bottleneck tendency of the load data, selects the countermeasure policy,
The management method according to claim 1. When the management computer determines that the tendency of the bottleneck load data is stable, the management computer selects a countermeasure policy candidate that reduces the load of the bottleneck resource over the entire period as the countermeasure policy. ,
The management method according to claim 2. When the management computer determines that the tendency of the bottleneck load data is not stable, the management computer selects a countermeasure policy candidate for reducing the peak value of the load of the bottleneck resource within the period as the countermeasure policy.
The management method according to claim 3. The management computer
A plurality of related load data are acquired for each of the plurality of related resources that affect the bottleneck resource among the plurality of resources .
The related load range is calculated based on each related load data, and when all the values in each related load data are within the corresponding related load range, it is determined that the tendency of the corresponding related load data is stable. ,
When it is determined that the tendency of the corresponding related load data is stable, the countermeasure means is selected based on the result of the determination of the tendency of the plurality of related load data.
The management method according to claim 4 . The first countermeasure measure candidate among the plurality of countermeasure means candidates indicates that the related resource in the bottleneck resource is moved to another resource of the same type as the bottleneck resource in the computer system.
In the management computer , the tendency of the bottleneck load data is determined to be stable, and the number of related load data determined to be stable among the plurality of related load data is not stable. When the number is less than the number of related load data determined to be, a countermeasure measure candidate other than the first countermeasure means candidate is selected as the countermeasure means from the plurality of countermeasure means candidates.
The management method according to claim 5 . The second countermeasure candidate among the plurality of countermeasure candidates indicates that the load on the related resource is reduced.
The management computer determines that the tendency of the bottleneck load data is not stable, and that the number of related load data determined to be stable among the plurality of related load data is not stable. When the number of related load data is less than the number of related load data, a countermeasure measure candidate other than the second countermeasure means candidate is selected as the countermeasure means from the plurality of countermeasure means candidates.
The management method according to claim 6 . When the bottleneck resource is a pool and the related resource is a VM (virtual machine), the first countermeasure candidate is to move the data of the VM to another pool.
The management method according to claim 7 . When the bottleneck resource is the pool and the related resource is the VM, the second countermeasure candidate is to limit the I / O amount of the VM.
The management method according to claim 8 . The management computer
For each tendency of the related load data, a plurality of countermeasures are determined by selecting the countermeasure means using the corresponding related resource based on the countermeasure policy and the tendency of the related load data .
For each trend prior SL-related load data, based on the security objectives, to determine the parameters of the corresponding protective means,
The management method according to claim 5 . It said management computer, and said is the determined determined measure means parameters, Ru is displayed on the display device,
The management method according to claim 1. Memory and
With the processor connected to the memory and computer system
With
The processor
Of the plurality of resources in the computer system, the bottleneck load data which is the time-series data of the load of the bottleneck resource in the period is applied to the bottleneck resource which is the cause of the performance deterioration of the computer system in the past period. Acquired,
For the related resource that affects the load of the bottleneck resource among the plurality of resources, the related load data which is the time series data of the load of the related resource during the period is acquired.
When the bottleneck load range is calculated based on the bottleneck load data and all the values in the bottleneck load data are within the bottleneck load range, the tendency of the bottleneck load data is stable. Judge,
When it is determined that the tendency of the bottleneck load data is stable, the performance degradation can be resolved from among a plurality of countermeasure policy candidates indicating the policy for determining the parameters necessary for operating the computer system. Select a countermeasure policy and
Based on the countermeasure policy and the tendency of the related load data, the countermeasure means for resolving the performance deterioration is selected from a plurality of countermeasure measure candidates indicating the operation of the computer system.
The parameters of the countermeasure means are determined based on the countermeasure policy.
Management calculator.

Images