JPS63136140A - Optimum job scheduling system - Google Patents

Abstract

PURPOSE:To improve job throughput by providing a data base for consumption amount of resources to decide the executing order of jobs to be processed based on the consumption amount of resources estimated from the data base. CONSTITUTION:A job processor contains a resources consumption amount data base 6, a resources consumption amount estimating means 7 and a job executing order control means 5A. In a job processing request mode a terminal 1 inputs a job processing request together with a key showing the job characteristics. Then the means 7 retrieves the data base 6 by means of said key to obtain an estimated resources consumption amount of the corresponding job. The means 5A totalizes the estimated consumption amount of resources shared by all jobs of a job processing queue 2 to evenly consume each shared resources based on the supply capacity and also the processing order is decided for the shared resources serving as critical paths so that these resources are carried out in the order of smaller power consumption amounts. Thus it is possible to shorten the waiting time for processing and to improve the job throughput capacity.

Description

[Detailed Description of the Invention] [Summary] In a job processing device that executes jobs using shared resources, a database of resource consumption is provided, and this is used to determine the number of jobs waiting to be processed based on the predicted resource consumption. Improve job processing ability by determining the execution order.

[Industrial application field]

The present invention relates to an improvement of a job scheduling method in a job processing device, and more particularly to an optimal job scheduling method that improves job processing capacity by determining the execution order of jobs waiting to be processed based on predictions of resource consumption. be.

2. Description of the Related Art In order for a job processing device that uses a computer to execute various types of processing (jobs) to exhibit its maximum processing capacity, job scheduling, which determines the order of the processing, must be performed optimally. In order to perform optimal job scheduling in this way, all shared $ including the CPU (Central Processing Unit) must be consumed by all sources (CPU, memory, channels, DASD (Magnetic Disk Device), LP (Line Printer), etc.). It is desirable to perform R-speed prediction and perform job scheduling based on this prediction.

[Conventional technology]

FIG. 4 shows a conventional job scheduling method. In the figure, 1 represents a terminal, 2 represents a job processing queue, and 3 represents a CPU and memory.

Shared resources such as channels, DASD, and LP, 4 are job execution means, and 5 are job execution order control means.

When a user requests a job from the terminal 1 to the job processing device, the user adds job class information, which roughly categorizes the amount of consumption of specific resources (usually CPU time), to the terminal 1. Enter from 1.

The job execution unit 4 executes the job while consuming the shared resource 3, but at this time, based on the control of the job execution order control unit 5, it prioritizes jobs from waiting jobs that require less CPU time according to the job class. Execute.

[Problem that the invention seeks to solve]

In the conventional job scheduling method, C
Since only information on PU time can be obtained, and quantitatively that information is only a rough guide to the class specified by the user, it is insufficient for effective use of resources, and therefore job scheduling There was a problem that it was not possible to perform optimization.

[Means for solving problems]

The present invention is intended to solve the problems of the prior art, and has the basic configuration shown in FIG. In a job processing device that executes a job in , a resource consumption data resource consumption database 6 stores data indicating resource consumption corresponding to job characteristics.

The resource consumption prediction means 7 searches the resource consumption database 6 based on the key indicating the job characteristics assigned at the time of the job processing request from the terminal 1 to obtain a predicted amount of resource consumption.

The job execution order control means 5A calculates the total predicted consumption amount of all shared resources of the jobs in the job processing queue 2, and calculates the average consumption of each resource based on the total value and the total shared resource supply capacity. At the same time, the execution order of jobs in the job processing queue 2 is determined so that jobs are executed in descending order of resource consumption for shared resources that become critical buses.

[For production]

When requesting job processing, the terminal inputs a key indicating the job characteristics, and the job processing device searches the resource consumption database using this key to obtain the predicted resource consumption amount of the corresponding job. Next, calculate the total predicted consumption of all shared resources of the jobs waiting to be processed, and consume each shared resource on average based on the supply capacity.For shared resources that are on the critical path, select the shared resources in order of decreasing consumed R amount. Since the processing order is determined and executed, the processing waiting time can be reduced and the job processing capacity can be improved.

〔Example〕

FIG. 2 shows one embodiment of the present invention.
The same parts as in the figure are indicated by the same numbers, and 5A is a job execution order control means, 6 is a resource consumption database, and 7 is a resource consumption prediction means.

The resource consumption database 6 is constructed as a database of recent job consumption results corresponding to job characteristics. In other words, the resource consumption amount of a job is generally determined by the processing type, editing/translation/linkage, and number of steps in the job.

In order to be able to use such data while waiting for job processing, we have created a resource consumption database 6. will be established.

When a user requests a job from the terminal 1 to the job processing apparatus, the user adds a key indicating the characteristics of the job and inputs the key from the terminal 1.

This key is determined by, for example, the processing type or the name of a similar job in the past.

The resource consumption prediction means 7 searches the resource consumption database 6 using this key, and obtains the predicted consumption of the shared resource corresponding to the job.

Next, the job execution order control means 5A calculates the total expected consumption amount for all shared resources of the jobs waiting to be processed in the processing queue 2, compares it with the processing capacity of the job processing device, and processes the job according to the following logic. determine the order.

■Determine the execution order of jobs in the processing queue so that each resource is consumed evenly.

■Resources that become critical buses will be dissipated.
Execute jobs in descending order of quantity.

The job execution means 4 processes the jobs in the job processing queue 2 in accordance with the order determined by the job execution order control means 5A.

In this manner, each time a job execution is completed, the actual resource consumption amount for that job is fed back to the resource consumption amount database 6 and recorded as data.

FIG. 3 is a diagram showing the time relationship of job processing between the method of the present invention and the conventional method, where fal indicates the conventional method and (bl
shows the method of the present invention.

For example, processing request #1 of a job with many DASD accesses.
#2. If #3 is waiting for processing, in the conventional method shown in Figure 3 (al), #1, #2, #3 with low CPU consumption
Since the jobs are executed in the order of , the CPU processing of job #3 is not completed at time t1 when all DASD processing is completed and the transition is made to CPU processing.

In contrast, in the method of the present invention, scheduling is performed by combining the length of the CPU processing time and the length of the DASD processing time so that both CPU and DASD resources are consumed on average. As shown, time t,
All jobs have been processed in the previous method, and it has been shown that the processing capacity is improved compared to the conventional method.

〔Effect of the invention〕

As explained above, according to the present invention, the consumption amount of all shared resources of jobs waiting to be processed is predicted based on the database covering all shared resources, and job processing is scheduled based on the supply capacity. It is possible to shorten the processing waiting time and improve the processing capacity of the job processing device. The method of the present invention can dynamically respond to changes in resource consumption over time, and it also makes quantitative predictions based on past results, making it possible to make fairly accurate predictions and reducing processing capacity. It is effective in improving

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the basic configuration of the present invention, FIG. 2 is a diagram showing an embodiment of the method of the present invention, and FIG. 3 is a diagram showing the time relationship of job processing between the method of the present invention and the conventional method. FIG. 4 is a diagram showing a conventional job scheduling method. 1-Terminal 2...Job processing queue 3--Shared resources 4...Job execution means 5.5A--Job execution order control means 6-Resource consumption database 7--Resource consumption prediction means

Claims (1)

[Claims] In a job processing device in which a job execution means (4) executes a job in a job processing queue (2) using a shared resource (3), data indicating an amount of resource consumption corresponding to a job characteristic. Prediction of resource consumption by searching the resource consumption database (6) based on the resource consumption database (6) that stores the resource consumption database (6) and the key indicating the job characteristics assigned at the time of job processing request from the terminal (1) Resource consumption prediction means (7)
) and the predicted total consumption amount for all shared resources of the job in the job processing queue (2), and from the total value and the total shared resource supply capacity, each resource is consumed on average, and the critical The job execution order control means (5A) determines the execution order of jobs in the job processing queue (2) so that the shared resources serving as paths are executed in order of decreasing resource consumption. Optimal job scheduling method.