JPH11120147A - Load distributed control method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an efficient distribution technology which controls the distribution to a computer that is in an overload state or a distribution technology which enables job distribution through the distribution control to a specific computer in the case of distributing plural kinds of jobs in a computer system consisting of plural computers. SOLUTION: In a computer system which has plural computer 107-1 to 107-3 and a means that mutually refers to load management information of the computers 107-1 to 107-3, a computer which is in a specific state that includes an overload state that happens when a load distributing means distributes plural kinds of jobs to each computer 107-1 to 107-3 and makes it execute them realizes the throwing control of new load and the overhead reduction of load distribution control by stopping the updating of the load management information for a fixed time that is preliminarily defined and separating itself from an object of load distribution control.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a load distribution control method in a computer system including a plurality of computers.

[0002]

2. Description of the Related Art In a large-scale system, it is becoming mainstream to conduct business on a cluster-type computer system constituted by connecting a plurality of computers as nodes.

In order to efficiently perform business processes on such a cluster type computer system, it is necessary to effectively utilize the computer resources of each node. For that purpose, processing must be distributed to each computer node so that business processing is not concentrated on a specific computer node.

As described above, in order to distribute the load among the computer nodes and execute the job efficiently, there are several methods for collecting the load of each computer and selecting a computer to be processed based on the information. Proposed.

Japanese Patent Application Laid-Open No. 8-287070 discloses that load information of a computer node is collected at regular time intervals in a shared memory accessible from each computer node or in a load information management table installed on a high-speed coupling device. Is used to select a computer. This load information is CP
U usage rate, I / O usage rate, main memory usage rate, and the like are listed.

Japanese Patent Application Laid-Open No. Hei 8-3055671 discloses that in a computer system composed of a plurality of computers interconnected by a network, load information is collected for each computer group which is a job allocation unit at the time of job allocation, and the load is the least. This is for selecting a computer group.

Japanese Patent Application Laid-Open No. Hei 9-167141 discloses that a resource use target value is defined in advance for each task, and a load collected in a resource use state management table installed on a high-speed connection mechanism that can be commonly accessed from each computer node. The information is compared with the above-mentioned resource use target value, and the most suitable computer for processing the job is selected. The load information includes a CPU usage rate, an I / O usage rate, and a main memory usage rate.

[0008]

In the above prior art, information on the entire target system is collected, and calculations are performed based on the collected information. That is, the computers belonging to the system must always operate the information collection function and continue to write the load information to the shared memory or the load information storage area installed in the high-speed coupling device which can be commonly accessed from the plurality of computers. Did not. Therefore, the overloaded computer had to continue to execute the above function.

Further, when a plurality of tasks are processed by the computer system, the information collection function must be continuously operated even when a state in which a specific task is desired to be performed temporarily on a specific computer node occurs. Was.

In consideration of efficient operation of the computer system, it is necessary to reduce the overhead of the operation of the load distribution function described in the above-mentioned problem and to improve the throughput of the business that should be performed.

Further, a computer in an overloaded state or in a state separated as a load distribution target is not subject to load distribution from the viewpoint of other computers, so that information other than the information that the load distribution function is stopped is not transmitted. No need to get. Therefore, when selecting a computer to execute a job, the amount of data to be handled is reduced by narrowing down and acquiring only the information of the computer on which the load balancing function is operating, and the overhead of the load balancing function is reduced, thereby reducing the tasks that should be performed. The throughput should be improved.

The present invention relates to a load distribution control method in a computer system including a plurality of computers, and has the following two objects.

A first object of the present invention is to reduce the overhead of the operation of the load distribution function, in particular, the overhead of the load distribution function for a computer that has been subjected to an excessive load, and to enable the planned distribution of business. An object of the present invention is to provide a load distribution control method.

A second object of the present invention is to provide an efficient load distribution control method by reducing the amount of data handled by the load distribution function, thereby reducing the processing and communication processing of the load distribution function.

[0015]

Means for Solving the Problems A plurality of computers executing a job and an area which can be written and referred to by each of the plurality of computers to store load information thereof are stored.
In a computer system including one high-speed coupling device, a shared extended storage device, a storage device of a specific computer, or another storage device, the plurality of computers collect load information of the own computer and store the load information of the entire system. By having a load information update processing function for updating the area and a computer selection function for selecting based on the stored load information when selecting a computer to execute a job, a job is input to a low load computer. In the load distribution control method for equalizing the load, in order to achieve the first object, the load information update stop condition of each computer placed in an area accessible from each computer as well as the area storing the load information. The time until the exclusion is released, which describes the suspension period of the load balancing function of the local computer when the above criterion and the exclusion specification criteria described above are satisfied A step of stopping the load information update processing function by information obtained from the area including, if the time until the negative release has elapsed provided a step of resuming the load information update processing function. The information collection stop condition allows not only load information for detecting an overload state but also information collection stop to be specified by time. Here, examples of the exclusion specification criteria include computer load information such as the upper limit of the CPU usage rate and the remaining amount of main memory usage, and combinations thereof.

[0016] In order to achieve the second object,
A step in which the computer selection function determines a computer that has stopped collecting information as a non-stopped computer based on the exclusion specification criteria, and a step in which the computer selection function selects a computer to execute a job from the non-stopped computer Are provided.

The step of stopping the load information update processing function of the computer based on the information obtained from the area in which the conditions for stopping the load information update processing function and the stop period are described allows the load on the computer in a specific state including overload to be reduced. The overhead due to the information update processing function can be reduced. If the information collection stop condition is specified by time, the computer system can be excluded from the load distribution with a specific computer depending on the time zone. As a result, the separated computer can be dedicated to a specific task.

The step of restarting the load information update processing function of the computer after the elapse of the suspension period allows the user to specify information collection restart according to the characteristics of the business.

A step of the computer selecting function discriminating, based on the information collection stop condition, a computer that has stopped the load information update processing function as a computer that has not stopped, and the computer selecting function executes a job from the computer that has not stopped. In the step of selecting a computer to be executed, only the information of the computer that is not stopped is acquired and processed, whereby the processing of the computer selection function can be made more efficient. Further, it is possible to suppress the job from being further input to the computer in the overloaded state.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described in detail with reference to the drawings.

First, a first embodiment will be described with reference to FIGS.

This embodiment is directed to a system constituted by one or more computers. It is assumed that a plurality of computers are connected to each other by means of a high-speed coupling device that holds an area that can be updated and referred to by each computer as a means for communicating with each other.

FIG. 1 shows a basic configuration of the present embodiment. The main storage device 10 managed by the OS 104 executed by the central processing unit 102 in the high-speed coupling device 101
It is assumed that a computer exclusion designation table 200, which is input information of the present invention, and a load information management table 105 that stores the load of each computer are installed on the computer 3. In the present embodiment, the excluded computer determination flag 106 is set in the load information management table 105.

The computer 107 connected by the high-speed coupling device 101 includes a central processing unit 108 and a main storage device 1.
09. The load distribution unit 300 that realizes the load distribution function is provided as a part of the OS 110, and is executed by the central processing unit 108. The load distribution unit includes a load information update processing unit 301 that implements a load information update processing function and a computer selection processing unit 302 that implements a computer selection processing function.

FIG. 2 shows the structure of the computer exclusion designation table 200. The computer exclusion designation table 200 includes an exclusion designation criterion indicating a condition for performing exclusion designation for each of the computers constituting the system, and an exclusion indicating an information update suspension time until the exclusion is canceled when excluded. It includes information for controlling the information update interval such as the time until release.

FIG. 3 is a detailed explanatory diagram of the load distribution unit 300. The load distribution unit 300 includes a load information update processing unit 301 and a computer selection processing unit 302. The load information update processing unit 301 collects the load information of the own computer at predetermined time intervals while referring to the computer exclusion designation table 200, and updates the load information of the own computer in the load information management table 105. When a computer selection request is issued from the scheduler 111, the computer selection processing unit 302 obtains information of the load information management table 105, determines the computer with the lowest load by referring to the excluded computer determination flag 106, The information is returned to the scheduler 111.

Next, the processing flow of the load information update processing unit 301 will be described with reference to FIG.

First, when the operation of the load distribution function is started in the system, an interval at which information is first updated is set (step 401).

Then, the following processing is continued until the operation of the load distribution function is stopped.

The information update process described below is performed at an information update interval defined in advance in step 401 (step 402).

The load information and time information of the own computer are compared with the exclusion specification criterion of the own computer in the computer exclusion specification table 200, and it is determined whether or not the exclusion specification criterion is satisfied (step 403).

If the above judgment is false, the own computer is not excluded from the load distribution, so the area for storing the load information of the own computer in the load information management table 105 is updated, and the processing is continued as it is. (Step 404).

If the above determination is true, the own computer is not subject to load distribution, so first, as in step 404, the area for storing the load information of the own computer in the load information management table 105 is updated ( Step 405).

Next, the excluded computer determination flag 106 is set (step 406).

Then, the information updating process is stopped until the time until the exclusion of the own computer defined in the computer exclusion designation table 200 elapses (step 407).

When the time until the exclusion is released has elapsed, the following processing is performed (step 408).

In order to indicate that the load is to be re-distributed again, the excluded computer determination flag is set (step 40).
9).

Next, the load information updating process is restarted (step 410).

The processing flow of the computer selection processing section 302 will be described with reference to FIG.

When the computer selection process is performed in response to a request from the scheduler or the like, the computer specified as an exclusion is first excluded from the load distribution target candidates by referring to the exclusion computer determination flag (step 501).

Next, the load information of the computers not excluded in step 501 is obtained from the load information management table 105, and the computer with the lowest load is selected (step 5).
02).

Finally, the information determined in step 502 is returned to the computer selection processing request source (step 503).

As described above, the first embodiment has been specifically described. In the above-described embodiment, the load information update process is executed when the information of the system exclusion designation table is obtained. The format to be read into the device 109-n may be used. In such a case, the contents of the system exclusion specification table are newly changed in the load distribution unit described in the above-described embodiment, and the contents of the system exclusion specification table are updated in the main storage device of the own computer. It is assumed that a part having the function of performing the above is newly provided.

Next, a second embodiment will be described. This is basically the same as the first embodiment, but different points will be described below.

In the above-described embodiment, the excluded computer determination flag 106 is newly installed to determine the computer whose operation of the load distribution function has been stopped. However, the above function is stopped based on the update time information of the load information management table. The computer that performed the determination may be determined. This eliminates the need to newly provide an excluded computer determination flag area.

FIG. 6 shows an overall configuration diagram in this case. Here, the load information management table 105 is stored in each computer 107-n.
Contains the date and time of the last update performed on the load information of the own computer.

The processing flow of the new load information update processing unit 301 will be described with reference to FIG. First, when the operation of the load distribution function is started in the system, an interval at which information is first updated is set (step 401).

Then, the following processing is continued until the operation of the load distribution function is stopped.

The information update process described below is performed at an information update interval defined in advance in step 401 (step 402).

The load information and time information of the own computer are compared with the exclusion specification criterion of the own computer in the computer exclusion specification table 200, and it is determined whether or not the exclusion specification criterion is satisfied (step 403).

If the above determination is false, the own computer is not excluded from the load distribution, so the area for storing the load information of the own computer in the load information management table 105 is updated, and the processing is continued as it is. (Step 404).

If the above determination is true, the own computer is excluded from the load distribution, and therefore, the computer exclusion designation table 2
The information update process is stopped until the time until the exclusion release of the own computer defined in 00 has elapsed (step 407).

When the time until the exclusion is released has elapsed, the following processing is performed (step 408).

The load information updating process is restarted (step 4)
10).

Next, the processing flow of the new computer selection processing unit 302 will be described with reference to FIG.

When a computer selection process is performed in response to a request from a scheduler or the like, first, the load information management table 10
5, the computer whose updating has been stopped is excluded from the candidates for load distribution (step 5).
04).

Next, the load information of the computer not excluded in step 504 is obtained from the load information management table 105, and the computer with the lowest load is selected (step 5).
02).

Finally, the information determined in step 502 is returned to the computer selection processing request source (step 503).

Next, a third embodiment will be described. This is basically the same as the first embodiment, but different points will be described below.

In the above embodiment, the description based on the stop state of the load distribution function may be replaced with a state in which the load distribution function is operated. In this case, an exclusion designation criterion for time designation is used. FIG. 9 shows a computer exclusion designation table described so as to operate only for the time when the load distribution function is desired.
Shown in Here, it is described that the computer 3 is to be subjected to load distribution for only two hours from 3:00 to 5:00.
A code without description indicating that the time until the exclusion is released is not stored.

The processing flow of the new load information update processing unit 301 will be described with reference to FIG.

First, it is determined whether or not the own computer is specified to start at a time in the computer exclusion specification table (step 411).

If the above determination is true, first, an exclusion computer determination flag is set (step 412).

Next, the processing is stopped until the start time defined in the computer exclusion designation table (step 413).

When the start time comes, the exclusion-computer discrimination flag is set (step 414).

When steps 412, 413, and 414 are completed, or when the determination in step 411 is false, the following processing is performed. First,
When the operation of the load distribution function is started in the system, an interval at which information is first updated is set (step 401).

Then, the following processing is continued until the operation of the load distribution function is stopped.

The information update process described below is performed at an information update interval defined in advance at step 401 (step 402).

The load information and time information of the own computer are compared with the exclusion specification criteria of the own computer in the computer exclusion specification table 200, and it is determined whether or not the exclusion specification criterion is satisfied (step 403).

When the above determination is false, the own computer is not excluded from the load distribution, so the area for storing the load information of the own computer in the load information management table 105 is updated, and the processing is continued as it is. (Step 404).

If the above determination is true, the own computer is not subject to load distribution, so first, as in step 404, the area for storing the load information of the own computer in the load information management table 105 is updated ( Step 405).

Next, the excluded computer determination flag 106 is set (step 406).

Here, in this embodiment, in order to cope with the suspension of the operation of the load distribution function, it is determined whether or not the time to release the exclusion corresponding to the exclusion criterion that has become true in the determination of step 403 (step 403). Step 415).

If the above determination is false, the operation of the load distribution function is stopped.

If the above determination is true, the following processing is continued. First, the information update process is stopped until the time until the exclusion of the own computer defined in the computer exclusion designation table 200 elapses (step 407).

When the time until the exclusion is released has elapsed, the following processing is performed (step 408).

In order to indicate that the load is to be re-distributed again, the excluded computer determination flag is set (step 40).
9).

Next, the load information updating process is restarted (step 410).

Next, a fourth embodiment will be described. This is basically the same as the first embodiment, but different points will be described below.

When the computer exclusion designation table is described with reference to FIG. 2 in the description of the above embodiment, the occupation amount of the main memory of the own computer is used as the exclusion designation criterion. , And both may be used. FIG. 11 shows an example of the computer exclusion designation table in this case.

As described above, in the embodiment of the present invention, the area for storing the load management information is held, and the device to which each computer is connected has been described as a high-speed coupling device. However, the device may be a shared extended storage device. . FIG. 12 shows a basic configuration in that case. All computers 1 that make up the system
It is assumed that the computer exclusion designation table 200 and the load information management table 105 are installed on the shared extended storage device 113 connected to the server 07-n.

Further, the device holding the above area may be a specific computer in a system having a storage device having an area that can be updated / referenced in common by each computer. FIG. 13 shows a basic configuration in that case. It is assumed that the computer 107-1 is located as a master computer, and the computer exclusion designation table 200 and the load information management table 105 are installed on the main storage device 109-1 of the computer 107-1. The area where the computer exclusion designation table 200 and the load information management table 105 are installed can be updated / referenced from all the computers 107-n constituting the system.

The device holding the above-mentioned area may be a disk device. The basic configuration in that case is shown in FIG.
Shown in It is assumed that a computer exclusion designation table 200 and a load information management table 105 are installed on a disk device 114 connected to all the computers 107-n constituting the system.

[0084]

According to the present invention, it is possible to reduce the overhead of the operation of the load distribution function of a computer that has become overloaded,
Efficient load distribution can be performed. By performing the target control of the load distribution at the designated time, planned business distribution can be performed.

Furthermore, since the load distribution function does not acquire load information of computers that are not subject to load distribution when the system is selected, the processing of the load distribution function and the communication processing are reduced, and the efficiency of load distribution is improved. Can be.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a computer system including a load distribution unit and a necessary data area for implementing a load distribution control method of the present invention.

FIG. 2 is an explanatory diagram of a system exclusion designation table serving as input information of the present invention.

FIG. 3 is a specific configuration diagram of a load distribution unit that realizes a load distribution function of the present invention.

FIG. 4 is a flowchart illustrating a process of a load information update processing unit included in the load distribution unit;

FIG. 5 is a flowchart illustrating a process of a computer selection processing unit constituting a load distribution unit;

FIG. 6 is another configuration diagram of a computer system including a load distribution unit and a necessary data area for realizing the load distribution control method of the present invention.

FIG. 7 is a flowchart illustrating a new process of a load information update processing unit included in the load distribution unit.

FIG. 8 is a flowchart illustrating a new process of a computer selection processing unit constituting the load distribution unit.

FIG. 9 is an explanatory diagram of a changed reference of a system exclusion designation table serving as input information of the present invention.

FIG. 10 is a flowchart illustrating a new process of a load information update processing unit included in the load distribution unit.

FIG. 11 is an explanatory diagram of a changed standard of a system exclusion designation table serving as input information of the present invention.

FIG. 12 is another configuration diagram of a computer system including a load distribution unit and a necessary data area for realizing the load distribution control method of the present invention.

FIG. 13 is another configuration diagram of a computer system including a load distribution unit and a necessary data area for realizing the load distribution control method of the present invention.

FIG. 14 is another configuration diagram of a computer system including a load distribution unit and a necessary data area for realizing the load distribution control method of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 101 ... High-speed coupling device 102: Central processing unit of high-speed coupling device 103: Main storage device of high-speed coupling device 104: OS of high-speed coupling device 200: Computer exclusion designation table 105: Load information management table 10
6 ... excluded computer determination flag 107-n ... computer
108-n: central processing unit of the computer; 109-n: main storage device of the computer; 110-n: OS of the computer;
300-n: load distribution unit, 301-n: load information update processing unit, 302-n: computer selection processing unit, 111-n:
Scheduler, 112-n ... business system.

Claims (7)

[Claims] 1. A computer system comprising a plurality of computers, wherein each of the plurality of computers collects the load of its own computer, and the collected information is commonly accessible from all of the plurality of computers. A load distributing control method for distributing a process operated by the plurality of computers based on the contents of the commonly accessible region by providing a step of disposing in a certain area; If the conditions for stopping are met, the steps of collecting and arranging the steps of arranging are stopped, and the computer of which the operations of the step of collecting and arranging are halted and the computers that are not stopped are From a computer that has not stopped the determining step, the collecting step and the arranging step Load sharing control method characterized by a step of selecting a computer to execute the job. 2. The load distribution control method according to claim 1, wherein the computer that has stopped the operations of the collecting step and the arranging step restarts the operations of the collecting step and the arranging step. A load distribution control method, comprising the step of restarting the operations of the collecting step and the arranging step when they match. 3. A computer system comprising a plurality of computers, wherein each of the plurality of computers collects a load on its own computer, and the collected information is commonly accessible from all of the plurality of computers. Providing a step of arranging in a certain area, and distributing the processing operated by the plurality of computers based on the contents of the commonly accessible area, wherein the load of the own computer exceeds a threshold value. In the case, the step of stopping the operation of the collecting step and the step of arranging; the step of determining the computer that has stopped the operation of the step of collecting and the operation of the step of arranging and the computer that has not been stopped; To select a computer on which to execute a job from computers that have not stopped the step of Load sharing control method characterized by having a flop. 4. The load distribution control method according to claim 3, wherein the computer that has stopped the operations of the collecting step and the arranging step restarts the operations of the collecting step and the arranging step. A load distribution control method, comprising the step of restarting the operations of the collecting step and the arranging step when they match. 5. The load distribution control method according to claim 3, wherein a CPU usage rate of the own computer is used as a load threshold value for stopping the collecting step and the arranging step. Load balancing control method. 6. The load distribution control method according to claim 3, wherein the use amount of the main memory of the own computer is used as a load threshold value for stopping the collecting step and the arranging step. And a load distribution control method. 7. The load distribution control method according to claim 1, wherein time information is used as a condition for stopping the collecting step and the arranging step.