JP3543658B2 - Job execution management system - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a job execution management system in a computer system, and more particularly to a job execution management system that automatically optimizes job execution management settings.
[0002]
[Prior art]
In job execution management of a computer system, a large number of jobs input to the system are classified into several job classes according to the type of processing, required memory capacity, processing time, urgency, and other characteristics. Queue). By extracting and executing the job based on the job multiplicity preset for each class, a variety of jobs can be efficiently processed in parallel one after another.
[0003]
In a conventional system, job management setting information such as the number of jobs (job multiplicity) that can be executed in parallel for each job class is given in advance by an operator, and the OS performs job execution management based on this information. Changing the job management information required operator intervention. A system in which a plurality of sets of job processing management information are stored in the OS in advance and the OS selectively uses them has been proposed (Japanese Patent Laid-Open No. 60-27949). All management information sets must be prepared by the operator.
[0004]
An attempt to eliminate the need for operator intervention is disclosed in Japanese Patent Laid-Open No. 2-42539 entitled "Automatic Job Scheduling Method". In this method, actual job processing status data for each job class is accumulated, and assignment of an initiator (job multiplicity) to each job class is actually performed based on status data such as the number of unprocessed jobs and average job execution time. Is determined so as to be suitable for the job processing status, thereby automatically improving the operation efficiency of the system according to the change in the job input status.
[0005]
[Problems to be solved by the invention]
According to this method, automatic optimization of the job multiplicity of a class can be performed based on past processing trends. However, in the conventional technology including this technology, no consideration is given to automation of class setting at all, so that the following problems remain.
[0006]
That is, as an extreme example, a queue that is classified into a class that assumes processing of 10 seconds or less and a class that requires more time requires a job that takes about 2 hours to process and finishes in several tens of seconds. It is assumed that a job to be executed is submitted. In this case, both jobs are registered in the same queue class (a class requiring more than ten seconds), but if a job of about two hours occupies the multiplicity of the class, it takes several tens of seconds. Even if the job is completed, the execution process may be delayed for a long time, which may be inefficient. It is also conceivable to prepare a large number of subdivided classes in advance, but depending on the actual job input situation, the majority of the classes will be useless, and omitting them will increase the efficiency of initiator assignment. In some cases.
[0007]
Thus, the present invention provides a job execution management system that creates and deletes a job queue class based on actual job input status data, thereby automating the job class setting and realizing optimal job execution management. The purpose is to:
[0008]
[Means for Solving the Problems]
In order to achieve this object, the present invention provides a job input unit for inputting a job to be input to a system into queues classified according to the scale thereof, and a job executing unit for extracting and executing a job from the queue. In a system for performing job execution management by setting the multiplicity of a job for each class, the system further includes a data creation unit for creating data for job execution processing, and a class based on the data of the data creation unit. And a change means for newly establishing a class in accordance with the judgment.
[0009]
As described above, it is determined whether or not the setting of the job class is appropriate according to the job input situation at that time, and if it is better to add the class, by providing a means for newly creating a job class, for example, It is possible to prevent a delay in job execution due to a plurality of jobs having different job scales being allocated to the same job class as in the related art, and improve job processing efficiency.
[0010]
Also, in the job execution management system according to the present invention, the data creation unit includes at least a unit for recording a waiting time of each job in the queue and a unit for recording an execution time of each job. The method is characterized in that execution time distribution data is created. By configuring the data creation means in this way, it is possible to understand which job has a large amount of waiting time in a given job input situation, and to determine the size of the job taking an irrational waiting time. Can be determined.
[0011]
Further, in the job execution management system according to the present invention, when the execution time of the job with a long waiting time is not appropriate for the job class to which the job is assigned, the determination unit may determine an appropriate class for the execution time of the job. Is determined to be newly established. As described above, jobs whose waiting time is longer than the job execution time (that is, the job size) are extracted, and whether or not the job class to which these jobs are assigned is appropriate (for example, a job that ends in several tens of seconds). Is not assigned to the same class as the job requiring two hours), and if it is not appropriate, a job class appropriate for the job scale is newly established to correct inefficient job execution management. be able to.
[0012]
Further, in the job execution management system according to the present invention, when the execution time of the job with a long waiting time is appropriate for the job class to which the job is assigned, the determination unit may further include a multiplicity for the job class. And the changing unit increases the multiplicity of the job class in accordance with the determination of the determining unit. When the system is configured in this way and the class of the job having a long waiting time is determined to be appropriate, the job execution multiplicity of the job class is increased to increase the waiting time of the job of the class. And efficient job execution management can be performed.
[0013]
The job execution management system according to the present invention further includes: a unit that records a job execution result of the job execution unit; a class deletion determination unit that determines deletion of the job class; and a determination based on the class deletion determination unit. A class deletion unit for deleting the job class having a low execution record. Even if a job class already exists, for example, a job class that has not been used at all for a long time is useless, and it is reasonable to delete the class itself and assign an initiator to another. There are cases. In such a case, if the execution results of the job class are monitored and the class is appropriately deleted, more efficient job execution management can be automatically performed.
[0014]
Further, the job execution management system according to the present invention is further characterized by further comprising display means for displaying the setting of the job queue class at that time to the operator. By displaying the effective queue class to the operator in this way, a criterion for selecting an appropriate queue for the job size estimated by the operator can be given.
[0015]
BEST MODE FOR CARRYING OUT THE INVENTION
Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 is a block diagram showing a configuration of an embodiment of a job execution management system according to the present invention. As shown in FIG. 1, this system includes a terminal device 1, a job queue class display unit 2, a job input unit 3, a queue (queue queue classified according to job scale) 4, a job execution unit 5, a job waiting time recording unit 6, a job execution time recording unit 7, a job size-based frequency distribution creating unit 8, a job queue class creation determining unit 9, and a job queue class creating unit 10. These components 1 to 10 have the following functions, respectively.
[0016]
The terminal device 1 performs a dialogue process with the user and relays communication with the server machine. The job queue class display means 2 displays the effective queue class on the server side on the terminal device 1 and gives a criterion for selecting an appropriate queue for the job size estimated by the user. The job input unit 3 allocates a job sent from the terminal device 1 to the queue 4. The queue 4 manages jobs waiting to be executed due to a shortage of resources such as a memory / disk or a high degree of multiplicity in a queue system. The job executing means takes out the jobs from the queue 4 and sequentially executes the jobs.
[0017]
The job waiting time recording means 6 records the waiting time of each job in the queue 4 on a job-by-job basis. The job execution time recording unit 7 records the execution time of each job in the job execution unit 5 in job units. These data are written to the log 14 shown in FIG. The job size-based frequency distribution creating means 8 aggregates the data of the log 14 and creates a distribution of the execution time at least for a job having a long waiting time.
[0018]
The job queue class creation judging means 9 pays attention to the job execution time having a large frequency in the result of the frequency distribution creating means 8, and the assigned class is appropriate for the job of the noted execution period (that is, the job scale). Is determined. If the execution time is appropriate in relation to the class of the queue, an instruction to increase the execution multiplicity in the queue is issued. On the other hand, if the job scale is not appropriate in relation to the class of the queue, an instruction to create a queue having a new class is issued. The job queue class creating means 10 creates and updates a queue according to this signal.
[0019]
The operation of the system configured as described above will be described below. FIG. 2 is a flowchart for explaining the operation of the system shown in FIG. First, the user selects a class that seems to be optimal for the job to be submitted (step S2). Classes correspond to queues. The job is input to the queue 4 by the job input unit 3 (step S3), and is taken out and executed by the job execution unit 5. At this time, the waiting time in the queue 4, the time required for the execution processing, and the assigned class (queue) are recorded in the log 14 by the job waiting time recording means 6 and the job execution time recording means 7 (step S4). This recording processing flow is repeated until the number reaches a preset N times in order to estimate the tendency of the job scale in the server.
[0020]
After the recording has been performed N times (step S1), the data is totaled based on the recorded logs (step S5). Here, for each class, jobs whose waiting time is longer than the job execution time are extracted based on a preset reference, and a frequency distribution of the execution times of these jobs is created. Based on the data created here, the job queue class creation determining means 9 determines whether to change the job execution management setting (step S6). That is, the presence or absence of a job execution time zone exceeding the frequency set in advance in the distribution is detected. If there is no job execution time zone, it is understood that the job execution management setting is appropriate at that time, and the log recording process is performed again without changing the setting. (Steps S1 to S4) are repeated.
[0021]
If there is a job execution time period exceeding the distribution frequency set in advance in step S6, it is further determined whether or not the execution time period matches the execution time period set in the class to which those jobs are assigned. Is performed (step S7). Accordingly, when there are a considerable number of jobs having a long waiting time with respect to the size, whether the class in which these jobs are submitted is appropriate for the job size, that is, the class setting for the job submission status Turns out to be appropriate. As a result, if the execution time zones match, it is determined that the class setting is appropriate but the processing capacity of the class is insufficient, and the job multiplicity for this class is increased (step S8). On the other hand, if the execution time periods do not match, it is determined that the class setting is not appropriate, and a class suitable for the job scale in the execution time period with a long waiting time is newly established (step S9). Finally, the job queue class is updated (step S10), and the latest information can be provided to the user by the job queue class display means 2.
[0022]
To give an example of a specific operation, if a job that finishes in several tens of seconds in a queue that assumes processing of about several tens of seconds is submitted beyond the multiplicity of the class, the result is equivalent to execution processing. You will have to wait for a period. In this case, although the waiting time is detected to be large, the job execution time matches the job execution time assumed for the queue, so that it is determined that the processing capacity for the queue is insufficient, and the multiplicity is increased. On the other hand, if a job that finishes in several tens of seconds is input to a queue that is assumed to be processed for about two hours, at least two hours will be waited if the queue is not empty. In this case, too, the waiting time is detected as being large, and the job execution time (about several tens of seconds) is different from the job execution time (about 2 hours) assumed in the queue, so that a new class suitable for a job of this scale is newly established. (A class that assumes a job of several tens of seconds is newly established.)
[0023]
FIG. 3 is a diagram showing a configuration of a further embodiment of the job execution management system according to the present invention. The system shown in FIG. 3 includes a job execution result recording unit 11 and a job queue class deletion determining unit 12 in addition to all the components shown in FIG. Further, the job queue class creating means 10 shown in FIG. 1 is replaced by the creating / deleting means 13 in the example of FIG. Except for these components, it is assumed that the configuration and operation are exactly the same as those of the embodiment shown in FIG.
[0024]
The job execution result recording unit 11 records / stores the job execution result in the job execution unit 5. Further, the job queue class deletion judging means 12 judges whether or not to delete a job queue class having a small execution result based on the data from the job execution result recording means 11. Specifically, the number of executions of each class is recorded for a predetermined period, and it is determined that a class whose execution result is 0 or less than or equal to a predetermined reference value is deleted. If the job queue class deletion judging means 12 judges that the class should be deleted, an instruction is issued to the job queue class creating / deleting means 13 to delete the class.
[0025]
As a result, a class that has no execution record and is no longer needed can be sorted out, and it is not necessary to consider the assignment of the initiator to the class, and the efficiency of job execution management can be improved. Note that the system according to the present invention enables a new class to be established as described above, so that even if a class that has not been executed is deleted, it can be newly established as needed.
[0026]
【The invention's effect】
As described in detail above, the job execution management system of the present invention records the tendency of the job execution time, and makes it possible to set the class according to the tendency at each time. As a result, not only optimization of the multiplicity for the class (queue) but also the optimization setting of the class can be performed dynamically and automatically, so that an irrational job execution waiting time can be reduced. Thus, the efficiency of job processing can be improved.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a job execution management system according to the present invention.
FIG. 2 is a flowchart illustrating an operation of the system shown in FIG. 1;
FIG. 3 is a diagram showing a configuration of another embodiment of the job execution management system according to the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Terminal device 2 Job queue class display means 3 Job input means 4 Queue 5 Job execution means 6 Job waiting time recording means 7 Job execution time recording means 8 Job scale frequency distribution creation means 9 Job queue class creation judgment means 10 Job queue class Creation means 11 Job execution result recording means 12 Job queue class deletion determination means 13 Job queue class creation / deletion means 14 Log

Claims (6)

Job input means for inputting jobs to the system into queues classified according to their scales, and job execution means for taking out jobs from the queue and executing the jobs, and setting the multiplicity of jobs for each class A job execution management system, the system further includes a data creation unit for creating data relating to the waiting time and the execution time of the executed job, and a necessity of creating a class based on the data of the data creation unit. And a setting change means for newly setting a class in accordance with the judgment. 2. The job execution management system according to claim 1, wherein the data creation unit includes at least a unit that records a waiting time of each job in the queue and a unit that records an execution time of each job. A job execution management system for creating data of job execution time distribution. 3. The job execution management system according to claim 2, wherein the determination unit determines that the execution time of the job having a long waiting time is not appropriate for the job class to which the job is assigned. A job execution management system characterized by determining that a class is newly established. 4. The job execution management system according to claim 2, wherein the determination unit further determines the job class if the execution time of the job with a long waiting time is appropriate for the job class to which the job is assigned. 5. A job execution management system, wherein the setting change unit increases the multiplicity of the job class in accordance with the determination of the determination unit. 5. The job execution management system according to claim 1, wherein said system further includes means for recording a job execution result by said job execution means, and a class deletion judgment for judging whether said class deletion is necessary. And a class deleting means for deleting the job class having a low execution record based on the judgment of the class deletion judging means. The job execution management system according to any one of claims 1 to 5, wherein the system further comprises a display unit for displaying the queue setting at that time to an operator.

Images