JPH07234847A - Scheduling method for job - Google Patents

Abstract

PURPOSE:To select proper allocation destinations for jobs when respective processors are different in processing capability by allocating the jobs to the respective processors on the basis of their calculated remaining capability. CONSTITUTION:A job 1 which is waiting to be executed, is registered in a job management table 220. When a processor use rate as the reference for alteration of algorithm for job scheduling is 40% at this time, both the use rates of processors 100-1 and 100-2 exceed the reference use rate. Consequently, a job scheduler 200a selects a processor having higher remaining capability, i.e., processor 100-2 as a processor to which the job 1 is allocated, and then allocates the job. When the processor use rate as the reference for the alteration of the algorithm is 60%, on the other hand, the job scheduler 200a selects the processor which has a low processor use rate, i.e., the processor 100-1 and allocates the job 1.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a job scheduling method in a computer system, and more particularly to a job scheduling method suitable for a computer system including a plurality of processors.

[0002]

2. Description of the Related Art In a computer system including a plurality of processors such as a multiprocessor system or a parallel computer system, it is important to use the computer system more effectively by increasing the parallelism of processing. . Therefore, for example, JP-A-1-283663
In the gazette, the usage rate of each processor constituting the computer system (called load factor in the above gazette) is calculated at fixed time intervals, and jobs are assigned to each processor according to the calculated usage rate. Accordingly, a method for equalizing the processor load is disclosed.

For example, there is a first processor having a low usage rate and a second processor having a high usage rate, and a first job having a high processor usage rate and a low I / O operation rate and a low processor usage rate are input. It is assumed that the second job having a large proportion of output operations is waiting for execution. In this case, in the job scheduler according to the method described in the above publication, the first job having a high usage rate is assigned to the first processor having a low usage rate. Further, the second job having a low processor usage rate is assigned to the second processor having a high usage rate. This equalizes the load on the first processor and the load on the second processor.

[0004]

The above-mentioned conventional technique is effective because the load can be properly distributed when the processing capacity of the plurality of processors constituting the computer system is uniform. However, when the processing capabilities of the processors that make up the computer system are not uniform, that is,
If the processing capacities of the processors are different, there are problems as described below.

The first problem is related to selection criteria of a processor for scheduling a job. For example, consider a case where the ratio of the processing capacities of the first processor and the second processor is 1:10, the usage rate of the first processor is 50%, and the usage rate of the second processor is 51%. In this case, according to the above-mentioned conventional technique,
The job scheduler selects the first processor having a low usage rate as the allocation destination of the newly generated job, and allocates this new job to the first processor. However, in this case, depending on the contents of the job, it is quite possible that allocating the job to the processor 2 having a high processing capacity will not widen the difference in usage rate between the two processors.

A second problem is selection of a job scheduling algorithm. The above-described prior art job scheduler schedules jobs based on only one scheduling algorithm of allocating jobs to processors with low utilization. For this reason, when the processing capacity of the processors configuring the computer system is not uniform, the above-mentioned problem occurs, and in some cases, appropriate job allocation may not be possible.

An object of the present invention is to provide a job scheduling method capable of selecting an appropriate job allocation destination when the processing capacity of each processor is different.

Another object of the present invention is to provide a scheduling method capable of more appropriate job allocation by changing an algorithm for job allocation according to the status of each processor constituting a computer system. To do.

[0009]

Means for Solving the Problems The inventors of the present invention have
Focusing on the point that the above-mentioned problems are caused by uneven processing capacities of the processors constituting the computer system, it was thought that the job should be scheduled in consideration of the processing capacities of the respective processors.

That is, in order to achieve the first object, the job scheduling method according to the present invention calculates the remaining capacity of each processor from the processing capacity and usage rate of each processor constituting the computer system, The feature is that the job is assigned to each processor based on the calculated remaining capacity. In a preferred aspect, the remaining capacity of each processor is calculated based on the result of measuring the time required for execution of a specific instruction sequence in each processor, and the remaining processing capacity and unused rate (1
The value obtained by subtracting the usage rate from) is calculated.

In order to achieve the above second object,
The job scheduling method according to the present invention calculates the remaining capacity of each processor from the processing capacity and the usage rate of each processor constituting the computer system, and as a result of the calculation, there is a processor with a predetermined usage rate or less. Or not. When this determination is positive, the processor to which the job is assigned is determined based on the usage rate of each processor. When the determination is negative, the processor to which the job is assigned is determined based on the remaining capacity of each processor.

[0012]

By selecting a processor to which a job is newly assigned based on the remaining capacity calculated by using the processing capacity and the usage rate of each processor, the remaining processing capacity of each processor is effectively used to schedule a job. Can be done.

Further, by changing the algorithm for allocating jobs according to the processing status of each processor, specifically, the usage rate, job scheduling can be realized according to the system status at that time. It becomes possible to do.

[0014]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a schematic configuration diagram of a computer system according to the first embodiment. In FIG. 1, 100-1 and 1
00-2 is a processor, 200a is a job scheduler, and 300 is a network that connects the processors 100-1, 100-2 and the job scheduler 200a. The job scheduler 200a includes a processor management table 210a, a job management table 220 that manages jobs waiting to be executed, and a job allocation program 230a. Further, each entry of the processor management table 210a corresponds to the processor 100-1 which constitutes the computer system.
And 100-2, CPU power, CPU
It consists of usage rate and remaining capacity fields. The job scheduler 200a includes processors 100-1 and 100-
It is realized by a processor connected to the network 300 in the same manner as 2. In the present embodiment, the job scheduler 200a is the processor 10 that executes a job.
0-1 and 100-2 are provided independently,
It is also possible to configure any of these processors to have a function as a job scheduler. Further, the number of processors that execute jobs is not limited to two, but two
As long as there are more than one, it does not matter how many exist.

FIG. 2 is a flowchart showing the flow of job scheduling processing according to the job allocation program 230a of the job scheduler 200a.

First, the job allocation program 230a
Requests the processors 100-1 and 100-2 to execute a predetermined specific instruction sequence (step 20).
00). The processors 100-1 and 100-2 execute the specific instruction sequence in response to the request from the job scheduler 200a (step 2010), and the processors 100-1 and 100-2 finish executing the instruction sequence. The job scheduler 200a is notified of the time required for the execution (step 2020). In the job scheduler 200a, the job allocation program 23
0a, the processing capabilities of the processors 100-1 and 100-2 are calculated from the reciprocal of the notified execution time of the instruction sequence. Then, the processing capability obtained for each processor is entered in the corresponding CPU capability field of the processor management table 210a (step 2030). For example,
If a predetermined specific instruction sequence sequentially executes 1 million instructions, and the processors 100-1 and 100-2 execute the instruction sequence in 1 second and 0.1 second, respectively,
The processing capabilities of the processors 100-1 and 100-2 are
Calculated as 1 MIPS and 10 MIPS, respectively.

On the other hand, each processor 100-1 and 100
-2 calculates the usage rate of its own processor at regular time intervals and notifies it to the job scheduler 200a (step 2040). The job scheduler 20 which receives the notification about the usage rate from the processors 100-1 and 100-2
0a writes the usage rate in the corresponding CPU usage field of the processor management table 210a (step 2).
050). Further, the job scheduler 200a obtains an unused rate (a value obtained by subtracting the usage rate from 1) of each processor by using the usage rate of each processor recorded in the processor management table 210a. Then, the remaining capacity of each processor is calculated by obtaining the product of this unused rate and the processing capacity recorded in the processor management table 210a. The calculated remaining capacity is entered in the corresponding remaining capacity field of the processor management table 210a (step 206).
0).

When an executable job is registered in the job management table 220, the job scheduler 200a refers to the processor management table 210a and finds a processor having a predetermined usage rate (for example, 40%) or less. It is determined whether or not (step 2070). If there is a processor whose usage rate is less than or equal to a predetermined usage rate, a processor with a low usage rate is selected and a job is assigned to that processor (step 2080). On the other hand, if there is no processor whose usage rate is less than or equal to the predetermined usage rate, a processor with a large remaining capacity is selected and a job is assigned to that processor (step 209).
0). The job is then executed on the assigned processor (step 2100). Job scheduling in the case illustrated in FIG. 1 will be specifically described by the above job scheduling method. In FIG. 1, as can be seen from the processor management table 210a, the processor 100-1 has a processing capacity of 1 MIPS and its usage rate is 50%, and the processor 100-2 has a processing capacity of 10 MIPS. However, the usage rate is 55%. At this time, the remaining capacity of the processor 100-1 is 0.5 MIPS and the remaining capacity of the processor 100-2 is 4.5 MIPS.
Is.

In this state, the job management table 220
It is assumed that job 1 is registered as a job waiting to be executed. Assuming that the processor usage rate, which is the standard for changing the algorithm for job scheduling, is 40%, the usage rates of the processors 100-1 and 100-2 both exceed the standard usage rate. Therefore, the job scheduler allocates job 1 by selecting a processor having a high remaining capacity as a processor to allocate job 1, that is, processor 100-2.

On the other hand, when the processor usage rate, which is the basis for changing the algorithm, is 60%, the job scheduler 200a selects the processor with a low processor usage rate, that is, the processor 100-1 and assigns job 1 to it. .

A standard value may be set for the processor usage rate, which is the basis for changing the algorithm.
The administrator of each computer system may be allowed to change the command.

According to the present embodiment, the job scheduling is performed by using the remaining capacity calculated by the job scheduler to provide the job scheduling method in consideration of the processing capacity of each processor.

Further, since the job scheduling algorithm is changed according to the situation such as the usage rate of each processor, according to the execution situation of the processing in each processor,
You can schedule jobs appropriately,
The computer system can be used more effectively.

FIG. 3 is a schematic configuration diagram of a computer system according to the second embodiment of the present invention.

In FIG. 3, 100-1 and 100-2 are processors, 200b is a job scheduler, and 300 is a job scheduler.
Is a network that connects the processors 100-1, 100-2 and the job scheduler 200b. In this embodiment, each processor 100-1 and 100-2
, State tables 110-1 and 110-2 are provided respectively. Further, the job scheduler 200b has a processor management table 210b, a job management table 220 for managing jobs waiting to be executed, and a job allocation program 230b, as in the case of the first embodiment. Processor management table 2
Each entry of 10b corresponds to the processors 100-1 and 100-2 which form the computer system, and C
It consists of fields of PU unused rate and remaining capacity. However, in this embodiment, the processor management table 210b is not provided with a CPU field recording the processing capability of each processor.

Next, the flow of job scheduling processing according to the job allocation program 230b in this embodiment will be described with reference to the flowchart of FIG.

First, each processor 100-1 and 10
0-2 execute a specific instruction sequence (step 400
0), the processing capacity obtained from the reciprocal of the execution time is entered in the CPU capacity field of each of the status tables 110-1 and 110-2 (step 4010). further,
Each of the processors 100-1 and 100-2 calculates the usage rate of its own processor at regular time intervals, and writes the obtained usage rate in the CPU usage field of each of the status tables 110-1 and 110-2. . In addition, the unused rate (value obtained by subtracting the used rate from 1) is obtained from the calculated utilization rate, and the remaining rate is the product of the unused rate and the processing capacity (CPU capacity) recorded in the state tables 110-1 and 110-2. Calculate ability and state table 1
Fill in the remaining capacity fields 10-1 and 110-2 (step 4020). At this time, the obtained usage rate and remaining capacity of the processor are respectively calculated by the job scheduler 20.
0b is notified (step 4030). The job scheduler 200b writes the CPU usage rate and remaining capacity of the notified processor in the corresponding entry of the processor management table 210b.

When an executable job is registered in the job management table 220, the job scheduler 200b refers to the processor management table 210b in the same manner as in the first embodiment, and the usage rate is the specified usage rate. It is determined whether the following processor is present (step 4050). If there is a processor with a specific usage rate or less, a processor with a low usage rate is selected and a job is assigned (step 4060). On the other hand, if there is no processor having a specific usage rate or less, the processor having a large remaining capacity is selected and the job is allocated (step 4070). The job is then executed on the assigned processor (step 4080).

As described above, according to the present embodiment, a job scheduling method can be provided in which the processing capacity of each processor is considered by scheduling the job using the remaining capacity calculated in each processor.

[0031]

According to the present invention, it is possible to provide an appropriate job scheduling method even when there are differences in the processing capacities of a plurality of processors that form a computer system.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a computer system according to a first embodiment.

FIG. 2 is a flowchart of a job scheduling process according to the first embodiment.

FIG. 3 is a schematic configuration diagram of a computer system according to a second embodiment.

FIG. 4 is a flowchart of job scheduling processing according to the second embodiment.

[Explanation of symbols]

100-i: processor, 110-i: state table, 20
0: job scheduler, 210: processor management table,
220: job management table, 230: job allocation program.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Toshiyuki Kinoshita 1099 Ozenji, Aso-ku, Kawasaki-shi, Kanagawa Ltd. System Development Laboratory, Hitachi, Ltd.

Claims (9)

[Claims] 1. A method for scheduling a job in a computer system comprising a plurality of processors, wherein the remaining capacity of each processor is calculated based on the processing capacity and the usage rate of each processor, and the calculation is performed. A job scheduling method, wherein a job is assigned to each processor based on the remaining capacity of each processor. 2. A job scheduling method in a computer system comprising a plurality of processors, wherein the remaining capacity of each processor is calculated based on the processing capacity and the usage rate of each processor, and the result of the calculation is calculated. It is determined whether there is a processor having a predetermined usage rate or less, and when the determination is affirmative, the processor to which the job is assigned is determined based on the usage rate of each processor, and the determination is negative. , The job scheduling method is characterized in that the processor to which the job is allocated is determined based on the remaining capacity of each processor. 3. The job according to claim 1, wherein the processing capacity of each processor is calculated based on an execution time obtained by executing a specific instruction sequence in each processor in advance. Scheduling method. 4. The job scheduling method according to claim 1, wherein the remaining capacity of the processor is calculated from the product of the processing capacity of the processor and the unused rate of the processor. 5. A job scheduling method in a computer system having a plurality of processors for executing a given job and a job scheduler for allocating the jobs to the plurality of processors, wherein each processor has its own utilization rate. And notifies the scheduler, and the scheduler receives the notification, calculates the remaining capacity of each processor from the processing capacity and the usage rate of each of the processors, and based on the calculated remaining capacity of each processor. And assigning a job to each processor. 6. A job scheduling method in a computer system, comprising: a plurality of processors executing a given job; and a job scheduler allocating the jobs to the plurality of processors, wherein each processor has its own usage rate. To notify the scheduler, receive the notification in the scheduler, calculate the remaining capacity of each processor from the processing capacity and the usage rate of each of the processor, the utilization rate among the plurality of processors Determines whether or not there is a processor having a predetermined usage rate or less, and when the determination is affirmative, allocates a job to each processor based on the usage rate of each processor, and the determination is negative. At some point, a job is assigned to each processor based on the remaining capacity of each processor. Scheduling method for a job, wherein a kick. 7. A job scheduling method in a computer system, comprising: a plurality of processors executing a given job; and a job scheduler allocating the jobs to the plurality of processors, wherein each processor has a respective processing capacity. The remaining capacity is calculated from the usage rate, the self remaining capacity calculated from each of the processors is notified to the job scheduler, and in the job scheduler that received the notification, based on the remaining capacity of each processor, A job scheduling method characterized by allocating a job to a processor. 8. A job scheduling method in a computer system, comprising: a plurality of processors executing a given job; and a job scheduler allocating the jobs to the plurality of processors, wherein each processor has a respective processing capacity. The remaining capacity is calculated from the usage rate, and the remaining capacity and usage rate of the processor calculated from each of the processors are notified to the job scheduler, and in the scheduler that has received the notification, the usage rate is predetermined. It is determined whether there is a processor below the rate, when the determination is affirmative, a job is assigned to each processor based on the usage rate of each processor, and when the determination is negative, each of the The job is assigned to each processor based on the remaining capacity of the processor. Method of scheduling jobs to be. 9. A job scheduling method in a computer system having a plurality of processors for executing a given job and a job scheduler for allocating the jobs to the plurality of processors, wherein the processing capacity of each processor is calculated in advance. The usage rate of each processor is calculated at a predetermined interval, and it is determined whether there is a processor whose usage rate is lower than a predetermined usage rate according to the existence of jobs to be assigned. If the result of the determination is positive, the processor to which the job is assigned is determined based on the usage rate. If the result of the determination is negative, based on the processing capacity and the usage rate. Job scheduling, characterized in that it determines a processor to allocate the job according to the calculated remaining capacity. Method.