JP2601693B2 - Automatic job scheduling method - Google Patents

Description

Description: FIELD OF THE INVENTION The present invention relates to job scheduling in a computer system, and more particularly, to automatic optimization of job scheduling.

[Conventional technology]

The nature of jobs that a computer system must handle varies widely. In order to efficiently and efficiently execute a large number of jobs having a variety of properties in succession in this manner, the jobs are divided into types of processing, required memory capacity, processing time, types and number of input / output devices used, and Jobs are classified into several job classes according to class, executable time, urgency, and other characteristics.Job scheduling, that is, management of job execution order, is basically based on the job class. Performed by In order to perform job scheduling, job processing management information that prescribes, for each job class, an assigned time slot, the number of jobs that can be executed in parallel, a priority, and the like is prepared in advance.

In conventional job scheduling, job processing management information is kept as it is when given to a computer by an operator, and changing it requires operator intervention. A job scheduling system in which a plurality of sets of job processing management information are provided and the OS selectively uses them, thereby improving the operation efficiency of the computer system has been proposed (Japanese Patent Application Laid-Open No. 60-27949). )But,
All of these job processing management information sets must be prepared by the operator.

[Problems to be solved by the invention]

Job processing management information prepared by the operator includes:
It is created based on the expected average distribution of the job class of the job to be input, while the distribution of the job class of the actual job is always fluctuating,
Frequently, deviations from the expected average distribution become so large that situations arise where it is no longer possible to represent an efficient job scheduling.

An object of the present invention is to automatically optimize job processing management information, thereby always realizing optimal job scheduling regardless of fluctuations in the job class distribution of jobs to be input.

[Means for solving the problem]

The automatic job scheduling according to the present invention accumulates job processing status data (for example, the number of jobs to be processed and average job processing time) for each job class, and assigns an initiator to a job class using the job processing status data. Updates the job processing management table so as to match the actual job processing situation. This updating can be realized, for example, by first calculating the required number of initiators of each job class, and then assigning the actually prepared initiators to the job class according to the calculated required number of initiators. . Also,
This update is preferably performed periodically or each time the required job amount for any of the job classes falls outside the predetermined range.

[Action]

Job processing status data that can be easily collected by the OS
It reflects the actual processing status of each job, and therefore, by updating the job processing management table using this data, the assignment of the initiator to the job class is
Adapted to the actual job handling situation. If there is a limit to the number of initiators that can actually be prepared, calculate the required number of initiators for each job class separately,
The method of assigning a real initiator according to the result is as follows.
It is valid. Further, the job processing management table is updated.
By performing the process periodically or every time the required job amount for any job class deviates from the predetermined range, the assignment of the initiator to the job class can always be kept optimal.

〔Example〕

FIG. 1 shows, in the form of a functional block diagram, an example of a computer system for executing job scheduling according to the present invention. In the figure, a wide arrow indicates the flow of job processing, and a single line arrow indicates a flow of control information.

The job input unit 1 receives successive jobs to be input to the computer system 10, classifies them into job classes, and registers them in the job queue 2. In Jobobukiyu 2,
Jobs form a queue for each job class.
An appropriate number of initiators 3 are provided, and each initiator takes out a job from the job queue 2 and sets its execution environment. The job scheduler 4 instructs the vacant initiator 3 from which job class queue the job should be removed. The job execution unit 5 executes the job extracted by the initiator 3 under the set environment, and the job output unit 6 performs an output process for the job that has been executed.

The job execution management unit 7 monitors the job waiting status in the job queue 2 and the job execution status in the job execution unit 5, and creates / updates a job processing status table 8 indicating the job processing status for each job class. As shown in FIG. 2, the contents of the job processing status table 8 in the present embodiment include the number of unprocessed jobs and the average job execution time (sum of CPU time and input / output time) for each job class.
Consists of

As shown in FIG. 3, the job processing management table 9 indicates to which job class each initiator is assigned to which time zone. In the figure, for example, the No. 2 initiator is assigned to job class B from 8:00 to 20:00, and is assigned to both job classes B and C from 20:00 to 8:00.
B is assigned in preference to C. If one initiator is assigned to more than one job class, the one listed above has the higher priority. The job scheduler 4 gives an instruction to the initiator 3 according to the job processing management table 9. The contents of the job processing management table 9 are automatically updated by the job execution management unit 7 using the contents of the job processing status table 8 in a procedure described later.

FIG. 4 shows, in the form of a flow chart, an example of the job scheduling process performed according to the present invention by the computer system shown in FIG. When the system is started (201), the initial settings (2
02) is performed. The job scheduler 4 instructs the initiator 3 to take out the job according to the job processing management table (207), and the job taken out by the initiator is executed by the job execution unit 5 (208). When the execution of each job is completed, the job execution management unit 7 observes the waiting status in the job queue 2 and the execution status in the job execution unit 5, and stores the status of each job class in the job processing status table 8 (FIG. 2). The number of processed jobs and the average job execution time are updated (211).

If the fixed time elapses or the number of unprocessed jobs of a certain job class becomes “0” while the above operation is repeated (203), the job execution management unit 7 uses the job process management table 8 Then, the number of initiators required for each job class is calculated (204). This calculation is performed according to the following equation.

X = n × m × α where X is the number of initiators required for a job class, n is the number of unprocessed jobs in the job class, m is the average job execution time of the job class, and α is the operation policy. Is a coefficient determined by
The value of α is the job class personality, machine performance, time zone,
Varies depending on other factors. For example, if α = 1, the required number of initiators of job class A in the situation shown in FIG. 2 is 42 × 0.08 × 1 = 3.36 → 4, that is, “4”.

Next, the number of initiators actually allocated to each job class is determined based on the required number of initiators of all job classes calculated as described above (20).
Five). If there is no limit on the number of initiators, the calculated required number of initiators can be used as it is. However, the number of initiators may be limited by the characteristics of the OS, and in such a case, correction is required. For example,
The maximum number of initiators that can be provided is allocated according to the ratio of the calculated required number of initiators. then,
Based on the determined number of initiators of each job class, the contents of the job processing management table 9 are updated (20).
6). A lower priority job class can share an initiator with a higher priority job class.
In the example shown in FIG. 3, the required number of initiators calculated in step 204 is adopted as it is, and four initiator Nos. 1, 3, 4, and 6 are assigned to job class A. Initiator No. 3 is assigned to both job classes A and B, and job class A is executed with priority.

Thereafter, the job scheduler 4 manages the initiator 3 according to the updated job processing management table 9. That is, when a certain initiator is released, the job scheduler 4 refers to the job processing management table 9 and refers to the job class assigned to the initiator in the time zone (if a plurality of job classes are registered, the highest priority is assigned to the job class). Start the initiator by specifying the job class that has the ranking). The above operation is repeated until an instruction to stop the computer is issued (209, 210).

By appropriately selecting the coefficient α used for calculating the required number of initiators, it is possible to set operating conditions unique to each job class. Further, various direct conditions may be instructed to the job execution management unit. For example, it is easy to set an upper or lower limit on the number of initiators of a certain job class.

Regarding the contents of the job processing status table, other information may be added, and instead of the number of unprocessed jobs and the average job execution time, other amounts that can be used as a guide for the required job amount and the average job processing time, For example, the job input count and the average job CPU time may be used.

Further, the job processing management table may be updated when the number of unprocessed jobs reaches a lower limit other than “0”, or may be performed when the number of unprocessed jobs reaches a predetermined upper limit. .

〔The invention's effect〕

As described above, according to the present invention, the assignment of the initiator to the job class is dynamically optimized according to the change in the actual job processing status, and as a result, an unexpected change occurs in the job input status. However, the operation efficiency of the computer system can be automatically and always kept at the optimum.

[Brief description of the drawings]

FIG. 1 is a functional block diagram showing one example of a computer system for executing automatic job scheduling according to the present invention.
FIG. 2 is a diagram showing an example of the contents of a job processing status table, FIG. 3 is a diagram showing an example of the contents of a job processing management table, and FIG. 4 is a diagram of the automatic job scheduling method of the present invention. It is a flow chart of one Example. 2 ... job queue, 3 ... initiator, 4 ... job scheduler, 5 ... job execution unit, 7 ... job execution management unit, 8 ... job processing status table, 9 ... job processing management table, 203 ~ 206 …… Job processing management table update processing, 211
… Job processing status table update processing.

 ──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Hiroshi Kurihara 1 Horiyamashita, Hadano-shi, Kanagawa Hitachi, Ltd. Kanagawa Plant (56) References JP-A-62-210546 (JP, A)

Claims (4)

(57) [Claims] A computer for classifying jobs successively input into job classes and operating a plurality of initiators for starting execution of these jobs in accordance with a job processing management table indicating assignment of the initiators to the job class. In the system, a step of storing job processing status data indicating a job processing status for each job class, and using the stored job processing status data to assign the initiator to the job class matches the actual job processing status. Updating the job processing management table so that the job processing management table is updated. 2. The automatic job scheduling method according to claim 1, wherein the job processing status data includes the number of required jobs and the average job processing time for each job class. 3. The job processing management information updating step according to claim 1, wherein the job processing management information updating step includes a step of individually calculating the number of initiators required for each job class, and a step of calculating an initiator which can be actually prepared by the calculated required initiator. Assigning the job class according to the number of jobs. 4. The job processing management information updating step according to claim 1, wherein the job processing management information updating step is performed at regular time intervals or whenever the job required processing amount for any one of the job classes deviates from a predetermined range. Automatic job scheduling method.