JPH04326434A - Precision improvement control method for job execution prediction - Google Patents

Abstract

PURPOSE:To highly precisely predict job execution time even when job execution environment varies by predicting the execution time of a job by means of the individual configuration elements and correcting execution prediction time through the use of processing history data. CONSTITUTION:The job processing history collecting part 14 of a job execution monitoring program 4 accumulates the passing time, the CPU time, the multiplicity and the number of the input/output issue times of an ended job 5 in an auxiliary storage device 2 with pseudo input/output time. A job scheduling part 10 prepares a predication table 6 based on a job processing history accumulated in the auxiliary storage device 2 at the time of starting a processing. A job execution prediction correcting part 12 periodically monitors the CPU time, the number of the input/output issue times, the passing time and the multiplicity of the job 5 so as to obtain correcting intermediate prediction execution time and corrects job finish schedule time in the prediction table 6 based on the result. Then the job start/finish time of an unexecuted job is corrected in accordance with the corrected content.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for predicting and controlling the job execution time of a computer system capable of executing at least one job, and more particularly to a method for predicting the execution time of a job that varies depending on the load on the computer system, and This invention relates to a method for correcting predicted execution time during execution.

0002

2. Description of the Related Art With the development of multiprogramming control functions of operating systems, methods for controlling job execution in computer systems that simultaneously execute a plurality of jobs are becoming common. However, when a large number of jobs are executed at once, competition occurs for the limited resources of the computer system, leading to a decrease in performance, which may cause the execution time to vary significantly even for the same job. As a solution to this problem, a method has been considered in which jobs with similar processing characteristics are grouped into one group (job class) and jobs are executed for each job class. According to this method, the number of jobs running simultaneously can be limited by the number of job classes to be executed, and resource conflicts can be avoided to some extent. A known example of job execution control using this method is “VOS3 Center Management, Hitachi HITA
C Manual, 8091-3-005. However, the problem with this method is that although it is possible to avoid resource conflicts to some extent by setting job classes and limiting the number of jobs that can be executed simultaneously, resource conflicts cannot be avoided because the job classes are specified manually. This means that the execution time of the job cannot be stabilized. Therefore, even if a job is being executed, it is not known when it will end. There are many jobs that must be completed at a predetermined time.

[0003] In order to solve this problem, it is necessary to predict the execution time of a job as accurately as possible, detect a delay state, and take countermeasures. To detect this delay state, the effective service rate (ESR) is determined by CPU time, input/output time, and elapsed time.
There is a method to detect and control this drop in the effective service rate. A publicly known example of this method is “VOS3 Center Management”
JSS4 Edition-, Hitachi HITAC Manual, 80
90-3-122''. This method makes it possible to avoid abnormal delays in job execution time.

0004

However, even with the above-described job effectiveness control method using the effective service rate, the execution time of each job cannot be accurately predicted because it is based on the effective service rate of the entire system. That is, C
The problem is that it cannot take into account the characteristics of individual jobs, such as whether or not parallel processing of PU processing and input/output processing is performed, so although it can be used to counteract processing delays in the entire system, it cannot be applied to control individual jobs. remain.

Therefore, an object of the present invention is to predict the execution time taking into account the load of the computer system and the characteristics of individual jobs, thereby making it possible to schedule jobs and take measures against execution delays based on the prediction results. It is about providing the means.

[0006]

[Means for Solving the Problems] In order to achieve the above object, the job execution prediction control method of the present invention includes a job execution record accumulation program, a job execution prediction control program, and a job execution prediction correction program in a computer system. , and a terminal that displays the predicted execution time of a job as an input/output device of the computer system.

[0007] To predict the execution time of a job using the job execution prediction control program, ESR shown in Equation 1, which is an index representing the processing speed of the job, is used.

[0008]

[Math 1]

[0009] ESR is a relative processing speed coefficient when the case where a job is executed alone is set to 1.0.
In order to approximate the value of SR to 1 divided by the multiplicity, a pseudo input/output time is used instead of the input/output time for the purpose of absorbing the characteristics of each job.

[0010] In the job execution record accumulation program, ESR
To calculate this, the CPU time, elapsed time, and number of input/output issues are measured and accumulated at the end of job execution. Furthermore, the multiplicity for determining the pseudo input/output time is also measured and accumulated. The pseudo input/output time (IOT) can also be determined by equation 2.

[0011]

[Math 2]

[0012] The job execution prediction control program calculates the predicted CPU time (YCPU), predicted pseudo input/output time (YIOT), and predicted multiplicity (YT) based on the execution results for each job accumulated by the job execution record accumulation program. Predict the expected number of input/output issues (YIO). The predicted execution time (YET) before execution of the job is calculated using Equation 3.

[0013]YET=(YCPU+YIOT)×YT
…(
Equation 3) The job execution prediction correction program calculates the elapsed time (CET) and elapsed CPU at any point during the execution of the job.
The time (CCPU) and the number of elapsed input/output issues (CIO) are measured, and the intermediate predicted execution time is determined using Equation 4.

[0014]

[Math 4]

However, in Equation 4, the accuracy of the intermediate predicted execution time immediately after the start of execution of the job decreases, so it is corrected as shown in Equation 5 and Equation 6. The accuracy of job execution prediction is improved by correcting this performance information during execution with the previously accumulated execution performance of the job. Corrected intermediate prediction execution time=YET×(1-R
) + Intermediate predicted execution time x R


...(Number 5)

[0016]

[Math 6]

[0017] Regarding the definition of ESR, please refer to the following publications.

"A method of performance management in computer systems and experiments using it, Journal of the Information Processing Society of Japan, Vol. 23, No. 6, pp. 591-598"

[0019]

[Operation] The job execution prediction accuracy improvement method of the present invention divides the execution time of a job that is executed regularly or repeatedly in a computer system into elements that are less susceptible to the execution environment. Prediction accuracy is improved by predicting each.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to FIGS. 1 to 4.

FIG. 1 shows the configuration of a computer system to which the job execution prediction accuracy improvement control method of the present invention is applied. The job execution prediction accuracy improvement control method of the present invention uses a central processing unit 1 including a main memory, which is a component of a conventional computer system, and an operating system that allows the central processing unit to simultaneously execute at least one or more jobs 5. In addition to a system 8 and a console 9 for operating the operating system 8, an auxiliary storage device 2 for storing job processing history and job schedule information, a terminal 3 for displaying job schedule information, execution status, and prediction results, and a job according to the present invention. It consists of a job execution monitoring program 4 equipped with an execution prediction accuracy improvement control method. In this embodiment, the job execution monitoring program 4 that implements the job execution prediction accuracy improvement control method of the present invention operates as one job under the control of the operating system 8; however, this program may be incorporated into the operating system 8. It is also possible to separate the accumulation of job processing history and the prediction program.

FIG. 2 discloses the detailed configuration of the job execution monitoring program 4 of the present invention. The job execution monitoring program 4 includes a job scheduling section 10 , a job execution prediction result display section 11 , a job execution prediction correction section 12 , a job activation processing section 13 , and a job processing history collection section 14 . The job execution monitoring program 4 includes a plurality of prediction tables 6 that store job prediction information and execution status. In this embodiment, the job execution monitoring program 4 is started as one job by inputting a START command from the console 9 in FIG. When the job execution monitoring program 4 is started, the operating system 8 first sends the job scheduling unit 10
control is passed to. The job processing history collection unit 14 receives control from the operating system 8 when the job ends.

FIG. 3 discloses the structure of the prediction table 6. The prediction table 6 includes a storage area 15 for job names to be scheduled, a storage area 16 for job start times or scheduled job start times, a storage area 17 for scheduled job end times determined from job start (scheduled) times and predicted job execution times, Estimated CPU time (YCP) determined before starting the job
U) storage area 18, storage area 19 for predicted pseudo input/output time (YIOT) calculated before starting the job, storage area 20 for predicted multiplicity (YT) calculated before starting the job, prediction calculated before starting the job Storage area 2 for input/output issue count (YIO)
1. Predicted execution time (YET) calculated using equation 3 before starting the job
) storage area 22 for the time that is the final limit of the job end time, storage area 23 for the time that is the final limit of the job end time, storage area 24 for the elapsed time from the start of the job currently being executed (CET), and elapsed CPU time (CCPU) from the start of the job currently being executed. , and a storage area 26 for the number of input/output issues (CIO) since the start of the job being executed.

The job processing history collection unit 14 in FIG. 2 collects the elapsed time, CPU time, multiplicity,
The number of input/output issues is determined and further stored in the auxiliary storage device 2 along with the pseudo input/output time determined using Equation 2. The job scheduling unit 10 stores the auxiliary storage device 2 at the start of processing.
A prediction table 6 is created based on the job processing history stored in the. Furthermore, the job execution order is determined based on the job schedule information stored in the auxiliary storage device 2 and the prediction table 6, and the scheduled job start time 16 and scheduled job end time 17 are stored in the prediction table 6. When the execution conditions for the job to be scheduled are met, the job activation processing unit 13 activates the job and updates the scheduled job start time 16 and scheduled job end time 17 in the prediction table 6.

FIG. 4 is a processing flow of the main part of the job execution prediction correction section 12. The execution prediction correction unit is configured to perform the job execution status monitoring terminal 3 at fixed time intervals or at the job execution status monitoring terminal 3 in FIG.
Otherwise, the process is triggered by a command input from the console 9. The processing unit 27 obtains the elapsed time of the job being executed, the elapsed CPU time, and the number of issued input/outputs from the operating system 8, and calculates the elapsed time (CE) of the prediction table 6, respectively.
T) 24, elapsed CPU time (CCPU) 25, and issued input/output count (CIO) 26 storage areas. Next, the processing unit 28 calculates the corrected intermediate predicted execution time using Equations 3, 6, and 5, and based on this result, the estimated job end time 1
Correct 7. Next, the processing unit 29 corrects the scheduled job start time 16 and scheduled job end time 17 of the unexecuted job, corresponding to the content corrected by the processing unit 28 for the job that has not yet been executed. In the processing units 30 and 31,
The scheduled job end time 17 and the job end time limit 23 in each prediction table 6 are verified, and when it is determined that the execution of the job will not end even when the job end limit time is reached, job delay processing 31 is executed. Job delay processing 31 includes displaying a delay on the job execution status monitoring terminal 3, reducing the TSS multiplicity, and contacting the system administrator.

The job execution prediction result display section 11 in FIG.
The job execution prediction result and the correction result by the job execution prediction correction unit 12 are displayed on the job execution status monitoring terminal 3 in the form of a time chart or the like.

[0027]

[Effects of the Invention] According to the present invention, by regularly monitoring the CPU time, number of input/output issues, elapsed time, and multiplicity of a job, it is possible to correct the job execution time predicted in advance. It becomes possible to inform the job submitter with high accuracy of the predicted end time of a job, which varies depending on the load of the computer system, and has the effect of clarifying the expected end time of a job based on conventional experience.

[0028]Furthermore, by highly accurate prediction of job end time, job execution delay status can be notified in advance.
You can implement some measures to improve the processing speed of the job,
This has the effect of localizing the impact on other jobs due to a delay in job execution compared to the conventional method.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of a computer system equipped with a job execution prediction accuracy improvement control method according to the present invention.

FIG. 2 is a diagram showing the structure of a job execution monitoring program that executes job execution prediction and displays and monitors the results.

FIG. 3 is a diagram illustrating the structure of a prediction table that stores job execution prediction data and progress data of currently executing jobs.

FIG. 4 is a processing flowchart of a job execution prediction correction unit that is the core of the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1...Central processing unit, 2...Auxiliary storage device, 3...Job execution status monitoring terminal, 4...Job execution monitoring program, 5...Job, 6...Prediction table, 7...Billing information, 8...Operating system, 9...Console, 10...Job scheduling unit, 11...Job execution prediction result display unit, 1
2...Job execution prediction correction unit, 13...Job activation processing unit,
14...Job processing history collection unit.

Claims (5)

[Claims] Claim 1: In an information processing system capable of executing at least one job, the processing history of a job that is currently being executed or has been executed is broken down into components of processing time and collected, and the execution time of the job is collected before execution of the job. A control method for improving job execution prediction accuracy, characterized in that the prediction is made for each component, and the predicted execution time of a job being executed is corrected using collected processing history data. 2. In the processing history collection, the CPU time consumed, the number of input/output issues, the elapsed time, and the multiplicity of the job are collected at any time during the execution of the job and at the end of the job; 2. The job execution prediction accuracy improvement control method according to claim 1, wherein the CPU time consumed, the number of input/output issues, the elapsed time, and the degree of multiplicity are stored. 3. In predicting the execution time, the expected CPU time, expected number of input/output issues, and expected pseudo input/output time of the job are set before execution of the job. The job execution prediction accuracy improvement control method described in . 4. In predicting the execution time, one of the jobs
A control method for improving job execution prediction accuracy according to claim 1, characterized in that the input/output processing time per run is determined in a pseudo manner using elapsed time, multiplicity, number of input/output issues, and CPU time. . 5. In the correction of the predicted execution time, the elapsed C
Using the PU time, elapsed time, and elapsed I/O issue count,
2. A job execution prediction accuracy improvement control method according to claim 1, characterized in that the progress rate is calculated and corrected.