JP5857144B2 - Virtual server management system - Google Patents

Description

  This invention is a server virtual that logically operates as multiple servers by setting up multiple OSs on a single server and allocating resources such as CPU, memory, and I / O of each server to each OS. Technology.

  In recent years, virtualization technology that enables a single server to be used as multiple independent servers instead of physically introducing multiple server computers has contributed to cost reduction, energy saving, and effective use of resources. As a thing, it attracts attention (refer nonpatent literature 1).

Specifically, when a program for realizing a virtual environment is introduced into a physical server and a plurality of OSs are set up on the server based on the functions of the virtualization program, each OS (hereinafter referred to as “virtual OS”) Server resources such as CPU, memory, and I / O are allocated.
As a result, each virtual OS functions as an independent virtual server, and by installing a separate application program for each, it is possible to achieve the same usability as if multiple servers were physically installed. It becomes possible.
5 minutes! Advantages of server virtualization Internet URL: http://www.atmarkit.co.jp/ad/hp/vse0701/vse01.html Search date: November 7, 2011

  In such a virtual environment, in order to truly realize the efficiency of processing, it is necessary to assign each job to which virtual server according to the characteristics of each job and the execution status of processing in each virtual server. Despite the importance of determining whether it is the most efficient at the present time, this has been neglected in the past.

  The present invention has been devised in view of such a current situation. According to the processing status of each virtual server, the optimal virtual server for executing each job is identified, and the virtual server is assigned to the virtual server. An object of the present invention is to provide a virtual server management technique capable of assigning jobs to the server.

  In order to achieve the above object, a virtual server management system according to claim 1 introduces a plurality of OSs into at least one physical server in which an OS is installed, and a virtualized environment constructed on the physical server. A virtual server management system comprising a plurality of virtual servers formed by the above, and the physical server and a management server connected to each virtual server, wherein the management server is of a specific type possessed by the physical server. Registers the resource allocation information storage means for registering the resource amount that is the amount of resources and the allocated resource amount allocated to each virtual server from the resource amount of the physical server, and the job to be executed Each virtual server based on the amount of resources held by the physical server stored in the job information storage means and the resource distribution information storage means By subtracting the total amount of used resources transmitted from the server, each virtual server is calculated from the processing for obtaining the current surplus resource amount of the physical server and the allocated resource amount of each virtual server stored in the resource distribution information storage means. By subtracting the amount of resource used above, the process of calculating the current amount of surplus resources for each virtual server, and when the amount of surplus resources of any virtual server exceeds the amount of surplus resources of the physical server The job information is calculated by applying resource utilization status calculation means for executing processing for correcting the surplus resource amount of the virtual server to the surplus resource amount of the physical server, and applying at least the surplus resource amount of each virtual server to a predetermined rule. The virtual server in charge of executing the job stored in the storage means is specified and the virtual server It is characterized in that a job assignment means for instructing the execution of the job to server.

  The virtual server management system according to claim 2 is the system according to claim 1, wherein the job assigning unit further selects a virtual server having the largest available resource amount among the virtual servers as the job. This is characterized in that it is specified as a virtual server in charge of execution of.

  The virtual server management system described in claim 3 is the virtual server management system according to claim 1 or 2, wherein a plurality of the physical servers exist and a plurality of virtual servers are formed on each physical server. When the virtual server is identified as the virtual server that is in charge of executing the job, the job allocating means determines that the virtual server belonging to the physical server having the largest spare resource amount is the virtual server in charge of executing the job. It is characterized by identification.

In the virtual server management system according to claim 1, since a marginal resource amount of each virtual server is obtained, and an execution destination of each job is specified based on at least the marginal resource amount. Therefore, it is possible to allocate a job to an optimal virtual server reflecting the current situation, and it is possible to realize processing efficiency.
Moreover, when calculating the amount of surplus resources of each virtual server, if the amount of surplus resources of a certain virtual server exceeds the amount of surplus resources of the physical server, a correction is applied to reduce this to the amount of surplus resources of the physical server. Because it has a mechanism, even if resources are allocated to each virtual server exceeding the physical resource amount of the physical server, jobs are executed in a state where the physical resource amount is exceeded in each virtual server. This can be effectively avoided.

  According to the virtual server management system of the second aspect, jobs are allocated to the virtual server having the most resources at the present time, so that efficient processing is realized.

  In the virtual server management system according to claim 3, when a plurality of virtual servers are respectively constructed on a plurality of physical servers, in addition to the surplus resource amount of each virtual server, the physical server to which each belongs Since the job execution destination is determined with reference to the amount of spare resources available, when multiple virtual servers with equivalent spare resources compete, the physical resources to which each belongs It is possible to accurately select a virtual server with a higher margin with reference to the server's spare resource amount.

FIG. 1 shows the overall configuration of a virtual server management system 10 according to the present invention, which includes a management server 12, a first physical server 14, and a second physical server 16.
The management server 12 and the first physical server 14 and the management server 12 and the second physical server 16 are connected via a communication network.
The number of physical servers is not limited to two, and it is naturally possible to connect three or more physical servers to the management server 12.

An OS (hereinafter “physical OS-A”) is set up on the first physical server 14, and an agent program (hereinafter “physical agent 18”) is installed on the physical OS-A.
In addition, by this physical agent 18, three virtual OSs (virtual OS-A1, virtual OS-A2, and virtual OS-A3) are activated on the physical OS-A. As a result, on the physical server 14, there are three virtual servers corresponding to the virtual OS-A1, virtual OS-A2, and virtual OS-A3.

An OS (hereinafter “physical OS-B”) is also set up on the second physical server 16, and a physical agent 18 is also installed on this physical OS-B.
In addition, three virtual OSs (virtual OS-B1, virtual OS-B2, and virtual OS-B3) are activated on the physical OS-B by the physical agent 18. As a result, on the second physical server 16, there are three virtual servers corresponding to the virtual OS-B1, virtual OS-B2, and virtual OS-B3.

The number of virtual OSs is not limited to three, and the physical agent 18 can naturally start four or more virtual OSs on a physical server.
Each virtual OS, like the physical OS, exhibits basic functions as an OS, but the physical OS and the virtual OS are not necessarily the same type of OS.
Each virtual OS has a virtual OS agent program (hereinafter referred to as “virtual agent 22”) installed therein.
Furthermore, various application programs necessary for executing business processing are installed on each virtual OS (not shown).

The management server 12 can individually command the virtual agent 22 to execute necessary jobs.
In addition, the physical agent 18 and each virtual agent 22 notify the management server 12 of data such as the usage status (utilized resource amount / actual measured value) of various resources and processing results as needed.

  FIG. 2 is a block diagram showing the functional configuration of the management server 12, and includes a resource usage status calculation unit 30, a job allocation unit 32, an execution result reception unit 34, a job allocation information storage unit 38, and a resource allocation information storage. Unit 40 and a job information storage unit 42.

The resource usage status calculation unit 30, the job allocation unit 32, and the execution result reception unit 34 are realized when the CPU of the management server 12 executes necessary processes according to a dedicated application program.
The job allocation information storage unit 38, the resource allocation information storage unit 40, and the job information storage unit 42 are provided in the external storage device of the management server 12.

As shown in FIG. 3, the resource allocation information storage unit 40 preliminarily defines resource allocation information held by each physical server.
For example, if the available amount of CPU resource of physical OS-A is expressed as 1000 in comparison with the CPU resource of the specified index OS, it is assigned to subordinate virtual OS-A1, virtual OS-A2, and virtual OS-A3. On the other hand, 400 each is assigned.
In addition, when the usable amount of CPU resources of the physical OS-B is 1200 in comparison with the index OS, 500 is allocated to the subordinate virtual OS-B and virtual OS-B2, and the virtual OS -400 is assigned to B3.
“CPU resource” means, for example, processing capacity per unit time.

Similarly, for the memory and I / O other than the CPU, resources held by the physical OS are allocated to the subordinate virtual OSs.
Also in this case, when the resource possessed by the index OS is set to 1000, the score obtained by comparison with the resource is expressed as the resource possessed by each physical server and the amount of resources allocated to each virtual server. It is expressed.
“Memory resource” means the memory capacity of the physical server.
“I / O resource” means, for example, the I / O speed of a physical server disk.

  Since it is rare for each virtual server to use all the resources allocated to it to the maximum at the same time, the amount of resources that are larger than the amount of resources owned by the physical server as described above is preliminarily assigned to each virtual server. Assigned. As a result, effective use of resources and efficient processing are realized.

The job information storage unit 42 stores the attributes of the execution target job.
FIG. 4 shows an example of data stored in the job information storage unit 42, and data of executable server, CPU, memory, I / O, maximum processing time, average processing time, execution status for each job. The item is set.

For example, JOB-1 defines virtual OS-A1, virtual OS-A2, virtual OS-A3, virtual OS-B1, virtual OS-B2, and virtual OS-B3 as executable servers.
Here, the executable server means a virtual server in which an application program for executing the job is installed.

In addition, for each executable server, a requested resource amount, maximum processing time, and average processing time for each resource of CPU, memory, and I / O are stored.
For example, when JOB-1 is executed on a virtual server of virtual OS-A1, CPU: 100, memory: 100, I / O: 100 resources are required, maximum processing time: 200 seconds, average processing time: When the same JOB-1 is executed on a virtual OS-B2 virtual server, while it takes 100 seconds, CPU: 150, memory: 150, I / O: 150 resources are required, and the maximum processing time is: It is specified that 300 seconds and average processing time are 150 seconds.
These data are set based on past results. At this time, the resources used in other virtual servers may be estimated from the results in a specific virtual server.

The “execution status” is a data item in which values representing the execution status of the job (“running”, “normal end”, “abnormal end”, “waiting execution”) are stored.
The value of this data item is updated as needed by the execution result receiving unit 34 and the job allocating unit 32. For example, if JOB-1 is a job that is executed every day at a predetermined time, a status of “waiting for execution” is set until the time arrives.
Next, when the time arrives and the job assignment unit 32 assigns JOB-1 to a predetermined virtual OS, a status of “in execution” is set.

Then, when the execution result (“normal end” or “abnormal end”) is transmitted from the virtual agent 22 of the virtual OS, the execution result receiving unit 34 sets a status corresponding to the execution result.
FIG. 5 shows the execution result data transmitted from the virtual agent 22 and includes at least data items of reception date / time, job name, server name, and execution result.

The data items of the job information are not limited to the above. For example, the requested resource amounts of the CPU, memory, and I / O can be set in detail by dividing them into maximum values and average values.
In addition to CPU, memory, and I / O, network resources (network usage status) can be added as resource types.

The job allocation information storage unit 38 stores a correspondence relationship between each job associated with the job allocation unit 32 and the virtual server.
FIG. 6 shows the job assignment information stored in the job assignment information storage unit 38, and includes at least data items of assignment date / time, job name, and server name.

Next, a processing procedure in the virtual server management system 10 will be described with reference to the flowchart of FIG.
First, the resource usage status calculation unit 30 generates resource usage status data and stores it in the memory (S10).
FIG. 8 shows an example of resource usage status data. For each physical OS and its virtual OS, time, CPU (total), CPU (used), CPU (capa), memory (total), memory (Used), memory (capa), I / O (total), I / O (used), and I / O (capa) data items are set at least.

In the “time” data item, a value indicating the passage of time is set. That is, the time “0” represents the current time, “+1” represents one minute later, and “+2” represents two minutes later.
However, the time interval is not limited to a unit of 1 minute. For example, “+ 1 = 5 minutes later”, “+ 2 = 10 minutes later”, “+ 3 = 15 minutes later”, etc. it can.

  In the data item of CPU (total), the total value (owned resource amount) of the CPU resource of the physical server is described in the case of a physical server, and the resource allocated to the virtual server in the case of a virtual server Value (allocated resource amount) is described. The value of this CPU (total) is acquired from the resource distribution information storage unit 40.

The CPU (used) data item indicates the amount of CPU resources that are currently used on each server (currently used resource amount) or the amount of CPU resources that will be used in the future (futurely used resource amount). A numerical value is described.
That is, with respect to the CPU (used) at time “0”, the measured value of the used resource amount transmitted from the physical agent 18 or the virtual agent 22 is filled.
On the other hand, with respect to the CPU (used) such as time “+1” or “+2”, the resource usage status calculation unit 30 sets a job information storage unit for each assigned job stored in the job allocation information storage unit 38. The estimated values calculated in the future by filling the respective requested resource amounts and the average processing time stored in 42 are filled. For example, even if three jobs with a requested resource amount “100” are assigned to a virtual server at the present time, and as a result, the CPU (used) at time “0” is 300, one of them If one job is completed by the time “+2”, “200” is filled as the CPU (used) of “+2”.

In the CPU (capa) data item, a numerical value indicating the CPU resource margin amount (room resource amount) in each server is described.
This CPU (capa) is simply obtained by “CPU (total) −CPU (used)” for the physical OS. The CPU (capa) of the virtual OS is also basically calculated by “CPU (total)-CPU (used)”, but this is a tentative value and is in balance with the CPU (capa) of the physical OS. In some cases, corrections may be made.

  For example, as shown in FIG. 9A, the physical OS-A CPU (capa) in a certain time zone is 300, the virtual OS-A1 CPU (capa) is 200, and the virtual OS-A2 CPU (capa) ) Is 100, and the CPU (capa) of virtual OS-A3 is 200, the CPU (capa) of each virtual OS is lower than the CPU (capa) of physical OS-A. Certified as a CPU (capa).

  On the other hand, as shown in FIG. 9B, when the CPU (total) of the virtual OS-A3 is 400 and the CPU (used) is 50, “CPU (total) −CPU (used)” The CPU (capa) required by this is 350, but this exceeds the CPU (capa) of the physical OS-A1 by about 50, so a correction is made to reduce it to 300.

The description of the data items related to the CPU described above also applies to each data item related to the memory and I / O.
That is, in the data item of the memory (total), in the case of a physical server, the total value (owned resource amount) of the memory of the physical server is described, and in the case of a virtual server, the data item is allocated to the virtual server. A memory distribution value (allocation resource amount) is described. The value of this memory (total) is acquired from the resource allocation information storage unit 40.

  The data item of memory (used) includes the amount of memory currently used in each server (currently used resource amount / measured value), or the amount of memory resource used in the future (future used resource amount / A numerical value indicating (estimated value) is described.

Also, in the data item of the memory (capa), a numerical value indicating a memory margin amount (a margin resource amount) in each server is described.
The memory (capa) is basically obtained by “memory (total) −memory (used)”, but for the virtual OS memory (capa), the rule is that the physical OS memory (capa) is the upper limit. Is applied, and any excess is reduced.

  In the case of a physical server, the I / O (total) data item describes the total value (owned resource amount) of the I / O resource of the physical server. The value of the I / O resource allocated to the server (allocated resource amount) is described. The value of this I / O (total) is acquired from the resource distribution information storage unit 40.

  The I / O (used) data item includes the amount of I / O resources that are currently used on each server (currently used resource amount / actually measured value), or I / O resource amount that will be used in the future. A numerical value indicating (future use resource amount / estimated value) is described.

Further, in the data item of I / O (capa), a numerical value indicating an I / O resource margin amount (margin resource amount) in each server is described.
This I / O (capa) is also basically obtained by “I / O (total) −I / O (used)”, but the I / O (capa) of the virtual OS is the I / O of the physical OS. Since the rule of O (capa) as the upper limit is applied, the portion exceeding this is reduced.

Next, the job allocation unit 32 generates job allocation information by referring to the resource usage status data (S12).
First, the job allocating unit 32 retrieves, in a predetermined order, jobs having a status of “waiting for execution” from the job information storage unit 42, and executes the job by referring to the resource usage status data. Among the possible servers, the virtual server that is currently most advantageous for executing the job is identified.
At this time, the job assignment unit 32 assigns each job to the optimum virtual server according to the following rules.

(1) Among the plurality of virtual servers, a server having the largest resource (for example, CPU (capa), memory (capa), I / O (capa)) at the present time is selected.
At this time, the type of resource to be prioritized for each job may be registered, and the virtual server having the highest margin resource amount related to the higher priority source may be selected.
For example, when the highest priority resource of job α is a CPU, this corresponds to a case where a surplus resource amount of other resources is ignored and a virtual server having the highest CPU resource surplus resource amount is selected.
Alternatively, it is possible to select a virtual server having the highest total amount of surplus resources for each resource, or to select a virtual server having the highest average value.

(2) Refer to the requested resource amount of each job, and select the virtual server with the largest available resource amount among the virtual servers that can currently provide the requested resource amount.
According to the rule (2), when there is no virtual server that satisfies the requested resource amount of each job, execution of the job is waited until the corresponding virtual server appears. On the other hand, when the rule (1) is applied, a virtual server having the largest available resource amount at the present time is selected without considering whether or not the requested resource amount of each job is satisfied. The

(3) Refer to the average processing time of each job and select a virtual server that can provide necessary resources over the time.
FIG. 10 exemplifies the relationship between the CPU (capa) of the virtual server and the requested resource amount of the execution scheduled job.
First, in the example of FIG. 10 (a), the CPU (capa) will decrease in about 5 to 13 seconds after the job is submitted, but it still exceeds the CPU resource amount required for the job to be executed. The job can be executed efficiently.
On the other hand, in the example of FIG. 10B, the CPU (capa) falls below the CPU resource amount necessary for the job in about 5 to 13 seconds after the job is submitted, and the resource is insufficient.
When such a resource shortage occurs in all the executable servers of a job, a virtual server that requires the shortest resource shortage time is selected.

(4) When multiple virtual servers are lined up under the same conditions, the amount of spare resources of the physical servers to which each virtual server belongs is compared, and a virtual server under a physical server with a larger amount of spare resources is selected.

The above rules (1) to (3) are applied alternatively, while the rule (4) is applied to the rules (1) to (3) in an auxiliary manner. .
However, these rules are merely examples, and the job allocation unit 32 can naturally select an optimal virtual server according to other criteria.
Further, which rule should be preferentially applied is defined in advance by a policy.

Next, the job allocation unit 32 transmits the generated job allocation information to the virtual agent 22 of the corresponding virtual server, instructs the execution of the job (S14), and stores the job allocation information in the job allocation information storage unit 38. (S16).
When the job execution result data is transmitted from the virtual server (S18), the execution result receiving unit 34 updates the “execution status” of the job information based on this (S20).

Until all necessary jobs are executed, the resource usage status calculation unit 30 updates the resource usage status data (S24), and the processing of S12 to S20 is repeated (S22).
This resource usage status data is updated at least at the start of job execution (at the time of submission) and at the end of execution of the job in order to update the surplus resource amount (capa) of each resource. Further, at the time of this update, the amount of used resources (used) in each server is measured.

When the number of virtual servers set in the physical server is large, the resources of the physical server are distributed to each virtual server and the resource shortage is likely to occur.
Therefore, it is effective to increase the substantial resources of the virtual server related to the target service by stopping the specific virtual server related to the other service.

  In the above, it has been assumed that the required resource amount of each job stored in the job information storage unit 42 is set in advance by humans, but it is effective to provide a mechanism for automatically calculating this. It is.

  Specifically, as shown in FIG. 11, a request resource amount setting mechanism 50 including a record information registration unit 52, a record information storage unit 54, and a request resource amount calculation unit 56 is provided in the virtual server management system 10. deep.

  The performance information registration unit 52 and the required resource amount calculation unit 56 are realized by the CPU of the management server 12 executing necessary processes according to a dedicated application program. The record information storage unit 54 is provided in the external storage device of the management server 12.

From the physical agent 18 and the virtual agent 22 to the actual result registration unit 52, the resource usage (actual value) of each job at each unit time, the resource usage (used) of the virtual OS that executed the job, Data such as the used resource amount (used) of the physical OS to which the virtual server belongs is transmitted.
The record information registration unit 52 stores these record data in the record information storage unit 54.

  When a certain amount of data is accumulated in the result information storage unit 54, the requested resource amount calculation unit 56 determines whether the virtual OS of the virtual OS at the time of job execution is based on the above-described result data and the resource distribution information in the resource distribution information storage unit 40. The amount of resource required (capa) is calculated, and the requested resource amount of each job is certified by comparing this with the amount of resource used by the job.

  For example, as shown in FIG. 12, the CPU resource usage amount of JOB-β is recorded with a score of 120 from time 0 to 3, during which the CPU (capa) of the virtual OS is stable at 200, Since a sufficient margin is maintained for the CPU resource usage amount of -β, the requested resource amount calculation unit 56 determines that “120” is appropriate as the requested resource amount of the CPU for JOB-β.

  On the other hand, the CPU resource usage amount of JOB-γ is recorded with a score of 100 from time 0 to 3, but the CPU (capa) of the virtual OS during that time is only 5, and the virtual OS side It shows that there was no room. For this reason, since the score of “100” is not reliable to qualify as the required resource amount of JOB-γ, the required resource amount calculation unit 56 determines that the JOB-γ uses the CPU resource usage amount. Make up. This is because if there are more CPUs (capa) on the virtual OS side, the amount of CPU resources used by JOB-γ may have increased further.

  In such a case, the CPU (capa) on the virtual OS side is increased, data is taken again, and the optimum required resource amount of JOB-γ may be derived.

  Judgment criteria of the requested resource amount calculation unit 56 described above, that is, how close the job resource amount is to the virtual OS's surplus resource amount is certified as "no margin", and how far away is the "surplus" certification Whether to do this is set in advance in the program of the requested resource amount calculation unit 56 as a threshold value.

It is a schematic diagram which shows the whole structure of the virtual server management system which concerns on this invention. It is a block diagram which shows the function structure of a management server. It is a figure which shows an example of resource allocation information. It is a figure which shows an example of job information. It is a figure which shows an example of execution result data. It is a figure which shows an example of job allocation information. It is a flowchart which shows the process sequence of this system. It is a figure which shows an example of resource utilization condition data. It is a figure which shows the relationship between CPU (capa) of a physical server, and CPU (capa) of a virtual server. It is a schematic diagram which shows the relationship between CPU (capa) of a virtual server, and the request | requirement resource amount of the execution scheduled job. It is a block diagram showing a configuration of a request resource amount setting mechanism that automatically calculates a request resource amount of each job. It is a figure for demonstrating the idea at the time of calculating the request | requirement resource amount of each job automatically.

10 Virtual server management system
12 Management server
14 First physical server
16 Second physical server
18 physical agents
22 Virtual Agent
30 Resource usage calculation part
32 Job allocation section
34 Execution result receiver
38 Job allocation information storage
40 Resource allocation information storage
42 Job information storage
50 Required resource amount setting mechanism
52 Performance Information Registration Department
54 Performance information storage
56 Required resource amount calculator

Claims (3)

At least one physical server with an OS installed;
A plurality of virtual servers formed by introducing a plurality of OSs in a virtualized environment constructed on this physical server;
A virtual server management system comprising the physical server and a management server connected to each virtual server,
The management server registers a resource amount that is the amount of a specific type of resource held by the physical server, and an allocated resource amount that is allocated to each virtual server from the resource amount of the physical server. Resource allocation information storage means;
Job information storage means for registering a job to be executed;
A process of obtaining the current margin resource amount of the physical server by subtracting the total value of the used resource amount transmitted from each virtual server from the resource amount of the physical server stored in the resource distribution information storage unit; A process of calculating a current margin resource amount of each virtual server by subtracting the used resource amount of each virtual server from an allocated resource amount of each virtual server stored in the resource allocation information storage unit; A resource usage calculation unit that executes processing for correcting the amount of surplus resources of the virtual server to the amount of surplus resources of the physical server when the amount of surplus resources of the virtual server exceeds the amount of surplus resources of the physical server;
The virtual server in charge of executing the job stored in the job information storage unit is specified by applying at least the margin resource amount of each virtual server to a predetermined rule, and the job is executed with respect to the virtual server. Job allocation means for commanding,
A virtual server management system comprising:   2. The virtual server management system according to claim 1, wherein the job allocating unit identifies a virtual server having the largest amount of spare resources as a virtual server in charge of execution of the job among the virtual servers. . The virtual server management system according to claim 1 or 2, wherein a plurality of the physical servers exist and a plurality of virtual servers are formed on each physical server.
When a plurality of virtual servers are identified as virtual servers in charge of executing the job, the job allocating means assigns a virtual server belonging to a physical server having the largest spare resource amount to a virtual server in charge of executing the job. A virtual server management system characterized by specifying a server.

Images