JPH02162436A - Collection system for system performance data - Google Patents

Abstract

PURPOSE:To facilitate the evaluation of the system performance by producing actually a resource using request to a control program which controls the resources necessary for execution of programs based on the resources and their using conditions and measuring the waiting time for use of the resources to output it after editing. CONSTITUTION:A work load setting means 3 inputs the resources necessary for execution of plural programs 1 using the resources and the using conditions of the resources via a directive 2 and sets them into a work load information table 4. A waiting time measurement means 5 measures the waiting time for use of the resources based on the information of the table 4 and by producing actually a resource using request and stores this measured waiting time into a measurement result store table 7. An editing output means 8 edits the stored waiting time and inputs this as a report 9. Thus it is possible to measure the waiting time for resources according to the necessary conditions without applying any load to the control program of an operating system. Then the waiting time can be analytically and simply obtained for the resources.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention relates to a system performance data collection method in a computer system, and in particular, when a program uses a resource, the waiting time from when a program issues a request until the resource becomes available. This relates to a method for measuring .

[Prior art] Evaluation of the performance of a computer system is essential when using it, and it is important not only to evaluate it from a theoretical perspective, but also to actually measure or collect various data that indicate the performance of the system (system performance data). and is used as evaluation material.

In conventional computer systems, control tables accumulated by the control program of the operating system,
The data on the management table was referenced and collected by a collection program (software monitor), and the data was edited and output.

The system performance data obtained includes resource usage time, resource usage, number of queue entries, etc., and the tI! of the usage status of each '\i source in the entire system. ! It was effective in detecting bottlenecks.

[Problem to be solved by the invention]

By the way, in a computer system equipped with a multiprogramming environment and in which various types of waiting may occur within the system, it is highly necessary to know the waiting time from when a resource usage request is issued to when the resource becomes available for each individual program. but,
In the conventional system performance data collection method described above, the waiting time of individual programs is not measured, and it is impossible to collect it as system performance data.

In other words, in order to obtain detailed wait information for each individual program using the operating system's control program, the control program must always be aware of this information, which increases the overhead of the operating system and reduces the amount of collected data. This is because an increase in the amount causes a significant drop in performance.

On the other hand, it is also possible to determine the waiting time from theoretical analysis, and the waiting time of resources in a simple system can be determined to some extent by using a queuing model, but it is not possible to obtain the waiting time for a resource in a simple system to some extent by using a queuing model. In large-scale computer systems that simultaneously support multiple types of processing, such as batch processing, various methods are used for servicing queues within the operating system, so resource waiting times can be easily calculated analytically. It was something I couldn't ask for.

The present invention has been proposed in view of the above points, and its purpose is to measure resource waiting times according to necessary conditions and collect the data without placing a burden on the control program of the operating system. The purpose of the present invention is to provide a method for collecting system performance data that can perform the following tasks.

[Means to solve the problem]

In order to achieve the above object, the present invention provides a method for collecting data indicating system performance in a computer system equipped with a multiprogramming environment and in which various types of waiting may occur within the system. a workload setting means for setting resources and usage conditions of the resources in a workload information table; and based on the information set in the workload information table, actually issuing a resource usage request to a control program that manages the resources. a waiting time measuring means for measuring a waiting time when using a resource and storing the measured waiting time in a measurement result storage table; and editing and outputting the waiting time stored in the measurement result storage table. The computer is also equipped with an editing output means.

[Effect]

In the system performance data collection method of the present invention, the workload setting means sets the resources required for program execution and the usage conditions of the resources in the workload information table, and the waiting time measuring means sets the resources necessary for program execution and the usage conditions of the resources. Based on the information set in the table, the waiting time when using the resource is measured by actually issuing a resource use request to the control program that manages the resource, and
Store the measured waiting time in the measurement result storage table,
&I aggregation output means edits and outputs the waiting time stored in the measurement result storage table.

〔Example] Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the system performance data collection method of the present invention. In FIG. 1, 1 is a plurality of programs that use resources (not shown), and 6 is an operating system control program that manages them. The functional parts in the present invention are workload setting means 3. A waiting time measuring means 51 and an editing output means 8,
Workload information table 4. Measurement result storage table 7
is a table that stores information necessary for processing. Also,
Directive 2 is input information, and report 9 is output information. The functions of each means are as follows.

Workload setting means 3; Directive 2 specifies the resources required for the execution of program 1 and the conditions for using those resources.
A function waiting time measurement means 5 inputs the information from and sets it in the workload information table 4; uses the resource by actually issuing a resource usage request to the control program 6 based on the information set in the workload information table 4; Function editing/output means 8 for measuring the waiting time and storing the measured waiting time in the measurement result storage table 7; Function for editing the waiting time stored in the measurement result storage table 7 and outputting it as a report 9. , the operation of the above embodiment will be explained with reference to other drawings.

FIG. 2 is a diagram showing the processing of the workload setting means 3. In FIG. 2, the workload setting means 3 inputs a directive 2 in a directive input step 31, analyzes the input directive 2 in a directive analysis step 32, and creates a workload information table 3.
Create.

The workload information table 4 includes information such as the resource 3 priority sampling frequency. Figure 3 shows an example of Directive 2, which is workload setting information.
Priority values are described in the order of measurement desired, followed by resources to be used (RESOURCE).

FIG. 4 shows the processing of the waiting time measuring means 5. In FIG. 4, the waiting time measuring means 5 obtains the next priority to be measured set in the workload information table 4 in a priority setting step 51, and sets it as the priority of its own program.

Next, in a resource selection step 52, information about the next resource set in the workload information table 4 to which a usage request should be issued is obtained.

Then, in a control program calling step 53,
The control program 6 that manages the resource is called by a normal supervisor call command or a direct call command, and a request to use the resource is issued. At this time, due to the difference in the control method of the control program 6, there is a type in which the request is registered in a queue and control is returned immediately, and a type in which a wait is made within the control program 6 and control is returned only after the request is satisfied. In the former case, a control program call step 53 waits for the completion of the request.
Do it inside.

Next, in the call result storage step 54, the control program 6 is stored in the control program call step 53.
Obtain the time from when a resource usage request is issued to when the resource becomes actually usable, that is, the waiting time, and calculate the resource name,
It is stored in the measurement result storage table 7 along with the priority.

Next, in step 55 for determining whether all resources have been processed, it is checked whether the waiting time measurement process for all resources stored in the workload information table 4 has been completed. If there are unprocessed resources, the process returns to the resource selection step 52. Also,
If all resources have been processed, the process proceeds to the next step (56) after initializing the processed resource pointer (not shown).

Next, in an all-priority processing completed determination step 56, it is checked whether the waiting time measurement processing for all the priorities stored in the workload information table 4 has been completed. If there is an unprocessed priority, the process returns to the priority setting step 51. In addition, if all the priorities have been processed, after initializing the processed priority pointer (not shown), the next step (5
Proceed to 7).

Next, in the sampling number determination step 57, it is checked whether or not measurements for the number of samplings stored in the workload information table 4 have been completed. If not completed, the process returns to the priority setting step 51. Further, if it is determined that the processing is completed, the processing of the waiting time measuring means 5 is terminated.

After the above processing, the editing output means 8 refers to the measurement results measured by the waiting time measurement means 5 and stored in the measurement result storage table 7, and sorts the same resource by priority as shown in FIG. 5(a). Waiting time - report 9 and/or waiting time for different resources (a, b, c, d) for the same priority as shown in FIG. 5(b)
Edit and output the current report 9.

[Effects of the Invention] As explained above, in the system performance data collection method of the present invention, based on the resources required for program execution and the usage conditions of the resources, the control program that manages the resources is actually By issuing a resource usage request, we measure the waiting time when using a resource and output it as [fi. It can be obtained in a short time and has the effect of making it easier to evaluate system performance. It is particularly effective in large-scale computer systems where various methods are used for queue service order within the operating system in order to simultaneously support multiple types of processing such as large-scale transaction processing, TSS processing, and hatch processing. It is.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the system performance data collection method of the present invention, FIG. 2 is an explanatory diagram of the processing of the workload setting means in FIG. 1, and FIG. 3 is a diagram showing the steps shown in FIGS. FIG. 4 is an explanatory diagram of the processing of the waiting time measuring means in FIG. 1, and FIG. 5 is a diagram showing an example of the report in FIG. 1. In the figure, ■...Program 2...Directive 3...Workload setting means 4...Workload information table 5...
・Waiting time measuring means 6... Control program 7... Measurement result storage table 8... Editing output means 9... Report

Claims (1)

[Claims] A method for collecting data indicating system performance in a computer system equipped with a multiprogramming environment and in which various types of waits may occur within the system. a workload setting unit that sets usage conditions in a workload information table; and a resource usage request that actually issues a resource usage request to a control program that manages the resource based on the information set in the workload information table. a waiting time measuring means for measuring the waiting time when the measured waiting time is to be used and storing the measured waiting time in a measurement result storage table; and an editing output means for editing and outputting the waiting time stored in the measurement result storage table. A system performance data collection method characterized by: