JPS603216B2 - Data processing system operation control method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a data processing system operation control method, in particular, a data processing system that periodically samples the load state of a data processing device by outside the hardware monitor, and based on the sampling result, The present invention relates to a data processing system operation control method that controls the amount of jobs executed by the data processing apparatus.

Data processing equipment that multi-processes multiple programs is generally considered to be operating under maximum load, but
However, in reality, it cannot be said that the system is always running at full capacity depending on the nature of each job.

For example, according to some types of analysis results, the rate of housewives working in real jobs is at most 70% to 80%.
%. '81 During job execution, the CPU load on which data processing equipment actually executes processing is 20% to 30% of the actual CPU load of the user task.

The multiplicity of 'C' processing is once 1.2 to 1.2 in terms of memory multiplicity. It is the degree of milk cliff.

Although it has not yet been fully analyzed what causes the operating conditions of the data processing equipment as described above, it is important to note that when looking at the operating conditions described above, It is desirable to adopt a means to actively promote job startup and apply a trigger from the outside to the master scheduler provided in the data processing device.

The present invention aims to solve the above-mentioned problems while correctly coordinating with existing processing systems. In the data processing device, the data processing device is provided with a job startup waiting table, a job execution waiting table, and a master scheduler that performs job startup processing and job execution waiting management by referring to the contents of both tables. a hardware monitor that periodically samples the operating status of the data processing apparatus, and a data processing equipment usage status analysis function unit that analyzes the operating status of the data processing equipment based on the sampling results from the hardware monitor. , the hardware monitor notifies the data processing device usage analysis function unit of the periodically sampled results, and the data processing device usage analysis function unit:
While referring to the notification, it is determined whether the busy status corresponding to the sum of the monitor program processing time and the user task processing time is larger or smaller than the target value, and the above-mentioned master scheduler is notified according to the result. The master scheduler is configured to issue a request including at least a hold request, a hold release request, and a job start request to the request, and upon receiving the request, the master scheduler issues the start waiting table and/or
Alternatively, the process corresponding to the above request is executed by referring to the job execution waiting table. This will be explained below with reference to the drawings.

FIG. 1 is an explanatory diagram showing the configuration of an embodiment of the present invention, and FIG. 2 is an explanatory diagram showing the operation control of the embodiment according to the present invention in the form of a flowchart.

In Figure 1, 1 is a data processing device represented by the information inside the main memory, and 2 is the hardware.
3 is a system management data processing device, 4 is a console, 5 is a master scheduler and is a control program that manages jobs for the data processing device operating under a multi-program, and 6 is a job start waiting table. 7 is a job execution waiting table that stores information about jobs that have completed their job stacks and are waiting to be started, in the form of a queue; 8 represents the CPU usage status information provided by the hardware monitor, and 9 represents the functions executed by the system management data processing device.

In a data processing device that executes a plurality of programs in multiprocessing, individual jobs that have completed a job stack and are waiting to be started are created in a queue using a job start waiting table 6 and kept waiting. The master scheduler 5 connects the job whose job stack has been completed to the queue of the table 6, and when it becomes possible to actually start a new job in the data processing reconfiguration, the job is placed at the top of the table 6, for example. The job will be extracted and started. In addition, illustrated table 6
Although this is a simple representation, in reality, the scheduler 5 has a pointer to, for example, table 6, and the contents of the pointer are stored at a predetermined location (A and A) that stores the job code of the job waiting to be started. The content of the specified location A is the specified location (B) that stores the job code of the next job that is waiting to be started along with the above job code.
-QUE link), and is configured as follows. Then, the scheduler 5 sequentially links the queue (
By following the QUE link) information, it is possible to know the job code of the job waiting to be started. Further, in the data processing apparatus, a job execution waiting table 7 is created, and jobs that have been started and are actually waiting to be executed due to processing circumstances are queued using the table 7 and placed in a waiting state. It will be done like this. That is, a plurality of jobs are sequentially executed for time-sharing processing, and the order can be considered to be dictated by the contents of the table 7 above. In addition, each queue connected to the table 7 has a priority of the task to which the job belongs for each job in the waiting state, and a hold display when the job is placed in the hold state (in the HOLD category). ) etc. are recorded. As shown in the figure, the master scheduler 5 uses the contents of the table 6 to process ■ job startup processing and register completed stacked jobs in the table 6 (create a queue), and ■ job Execution queue management process 0 Executes processes such as CPU allocation management processing for the job to be executed next and hold management processing to be described later.

In addition, the master scheduler 5 in the figure is the data processing device 1.
Although it is expressed as existing directly above the main memory bag of
Needless to say, in the case of a multiprocessor system, the system management processor 3 may be considered to have the functions. The system executes jobs one after another using the tables 6 and 7 and the master scheduler 5 as described above, but in the case of the present invention, the situation in which the data processing device 1 is actually operating and executing processing is determined by the hardware.・Monitor 2
It is checked by. Although the hardware monitor 2 exists as a known device,
In one embodiment of the present invention, the operating status of the data processing device 1 is grasped, for example, in the following manner. That is, while the data processing device 1 is executing processing, a flag indicating this is set when processing related to the monitor program is currently being performed, and when processing related to the user task is currently being performed. A flag is set to indicate this. hardware·
Monitor 2 samples the above flags at intervals of several ms,
Statistics of the situation at the time of sampling are taken, and the number of times the data processing device 1 was in a busy state (BUSYTIME)
, and the number of times it was in the waiting state (WAITTIME) are reported to the system management data processing device 3, respectively. Based on the above report, the system management data processing device 3 monitors at intervals of, for example, 50 hs or 1000 mS.

Analyzing whether the ratio of the time during which the data processing device 1 is in a busy state between the program processing time and the user task processing time is larger than, smaller than, or approximately equal to the target value,
■If the above ratio is large, issue an instruction to the master scheduler 5 to repeatedly hold jobs related to predetermined tasks for arbitrary predetermined periods, or if the above ratio is approximately appropriate. When it returns, it issues an instruction to release the above-mentioned hold repetition, ■ If the above-mentioned ratio is small, it issues an instruction to request the master scheduler 5 to start the job, and ■ In addition, as necessary, the above-mentioned In order to enable job startup, it issues instructions to the operator to stack jobs. Each of the above instructions ■, ■, ■ is issued to the master scheduler 5, and the master scheduler 5 is activated in response to each of the above instructions, and based on the contents of the table 6.7 above, Process.

That is, in the case of a hold instruction, the queue related to Table 7 is examined, for example, a job belonging to the task that uses the data processing device most frequently is extracted, and a hold instruction flag is set on the hold classification shown in the figure. Do it like this. By doing this, for example, when the job enters the execution state, the job is made to wait for one time, and the next time it enters the execution state, it is executed and the hold instruction flag is set again. You will be able to do it. Further, when a hold release instruction is given, the hold instruction flag is cleared. Further, in the case of a job start instruction, the scheduler 5 is configured to immediately start the job at the head of the queue based on the contents of the table 6. The above instruction ■ is displayed on the console 4, for example,
If there is an unstacked job, the operator is instructed to stack the job and make it ready for activation. FIG. 2 shows the above process in the form of a flowchart.

The process A shown in the figure corresponds to the operation status extraction by the hardware monitor 2 shown in FIG.
Processes 1 and J correspond to the processes of the master scheduler 5, and illustrated process K corresponds to the display on the console 4. As explained above, according to the present invention, job execution is controlled according to the operating status of the data processing device, and the execution of the job is controlled based on the information extracted by the hardware monitor. The execution of jobs is passively controlled. Therefore, the operating status of the data processing device is maintained substantially close to the predetermined target value. Note that the meaning of "passive" in the above is that it corresponds to the spontaneous activation of the master scheduler 5, and the data processing device 1 itself corresponds to a known hardware monitor. This does not preclude issuing the above-mentioned instructions by providing a hardware monitor or the illustrated function 9 as monitor information extraction means.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram showing the configuration of an embodiment of the present invention, and FIG. 2 is an explanatory diagram showing the operation control of an embodiment of the present invention in the form of a flowchart. In the figure, 1 is a data processing device, 2 is a hardware monitor,
3 is a system management data processing device that corresponds to the data processing device usage analysis function unit referred to in the present invention; 4;
5 is a console, 5 is a master scheduler, 6 is a job start waiting table, and 7 is a job execution waiting table. 2 diagrams

Claims (1)

[Claims] 1. In a data processing system having a data processing device that multi-processes multiple programs, the data processing device performs job startup processing and job execution by referring to a job startup waiting table, a job execution waiting table, and the contents of both tables. It includes a master scheduler that manages execution queues, a hardware monitor that periodically samples the operating status of the data processing device, and a hardware monitor that monitors the operating status of the data processing device based on the sampling results of the hardware monitor. The hardware monitor notifies the data processing device usage analysis function section of the periodically sampled results, and the data processing device usage analysis function section With reference to the said notification,
Determine whether the busy status corresponding to the sum of the monitor program processing time and the user task data processing time is larger or smaller than the target value, and depending on the result,
The master scheduler is configured to issue a request including at least a hold request, a hold release request, and a job start request to the scheduler, and upon receiving the request, the master scheduler refers to the start wait table and/or the job execution wait table. A data processing system operation control method characterized in that a process corresponding to the above request is executed.