JP2005209077A - Job scheduling method, computer system, and job scheduling program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an easy and effective scheduling method which suppresses use of resources of a system in emulation or the like without burdening a user in a job execution system capable of executing a plurality of jobs in parallel. <P>SOLUTION: When a job scheduling means selects a job to be next executed, from a job queue holding jobs waiting for execution, the number of jobs being executed at present is examined by a job execution means with respect to each job name, and a job of a job name having no jobs being executed at present or the smallest number of jobs being executed at present is selected. Thus job names of simultaneously executed jobs are distributed as much as possible. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a job scheduling method for controlling the job execution order in a computer system, and more particularly to a job execution scheduling method in a computer system provided with means for executing a plurality of jobs in parallel.

  In a system that can execute jobs in parallel, shared resources such as the central processing unit (CPU) and I / O channels are limited in execution, and the resources that are used differ depending on which resources are used. Job scheduling is necessary for efficient use. In the conventional job scheduling method, the user inputs resource usage prediction information into the system, analyzes the job to be executed in advance and performs emulation, predicts resource usage, and schedules jobs according to the amount of resources used, etc. (See, for example, Patent Document 1).

JP-A-9-311795 (paragraphs 0009-0015)

  However, in the method in which the user inputs the resource use prediction information to the system, there is a problem that the user is burdened with the resource use prediction, and the prediction is only a guideline and lacks accuracy. Further, the method of predicting resource usage by performing job analysis or emulation has a problem that time and resources are required for the work itself.

  SUMMARY OF THE INVENTION An object of the present invention is to provide a job scheduling method, a computer system, and a job scheduling program that do not require input by a user and that can perform job scheduling simply while suppressing the use of system resources. To do.

  The job scheduling method according to the present invention is characterized in that in a computer system capable of executing a plurality of jobs in parallel, scheduling is performed according to the job name of the job being executed.

  In the scheduling method, for example, when a new job can be executed, a job having a name that matches the name of a job that has the same job name with a small number of jobs is preferentially executed.

  The scheduling method may be configured to use a table in which the number of jobs being executed and waiting is set according to job names.

  The computer system according to the present invention schedules job execution means for executing a plurality of jobs in parallel, a job queue for holding a waiting job, and a job name of the job being executed, and is held in the job queue. Job scheduling means for passing a job to job execution means is provided.

  When the job execution unit becomes able to execute a new job, the job scheduling unit selects a job having a name that matches the name of the job that is being executed by the job execution unit with a small number of jobs having the same job name. Alternatively, it may be configured to take out from the job queue and pass it to the job execution means.

  When the job scheduling means has a table in which the number of jobs being executed by the job execution means and waiting in the job queue is set according to the job name, and the job execution means becomes capable of executing a new job The job having the name of the job corresponding to the job having the smallest number of jobs in execution in the table may be passed to the job execution means. In that case, the job queue is divided according to the name of the job.

  If the job scheduling means receives a job queuing request and the job name of the job does not exist in the table, the fields corresponding to the job name (job name, number of jobs being executed, number of jobs waiting) When the end of the job is notified from the job execution means, the number of jobs being executed in the column corresponding to the job name of the job is decremented by 1, and the number of jobs after subtraction is 0. In such a case, it is preferable that the column is deleted.

  According to the job scheduling program of the present invention, a job execution unit can execute a new job on a computer including a job execution unit that executes a plurality of jobs in parallel and a job queue that holds a waiting job. Sometimes, a job having a name that matches the name of a job with a small number of jobs having the same job name among jobs being executed by the job execution unit is extracted from the job queue and passed to the job execution unit. And

  According to the present invention, there is an effect that job scheduling for efficient use of resources can be performed without consuming the user's burden and using less system resources.

Embodiment 1 FIG.
Next, embodiments of the present invention will be described with reference to the drawings. Referring to FIG. 1, a computer system that executes the job scheduling method according to the first embodiment of this invention exists in a job queue 102 that holds jobs to be executed (waiting jobs) and the job queue 102. A job scheduling unit 101 for selecting a job to be executed next from a job, a job execution unit 103 for executing a job, an input / output resource 105 used for job execution, and a computing resource (hereinafter referred to as a computing resource). .) 106 and a memory resource 107. The job execution unit 103 includes a plurality of job execution units (Threads) 104 that actually execute jobs. Jobs to be executed are entered into the job queue 102 from the user or the job being executed.

  The input / output resource 105 is specifically an input / output channel in the computer system. Specifically, the calculation resource 106 is a central processing unit (CPU) in a computer system, or hardware or software having a structure specialized for a specific type of operation. Specifically, the memory resource 107 is a magnetic disk or the like in the computer system. The job execution unit 103 is, for example, a job management program that performs job execution management (management such as activation, suspension, and termination). Each job is managed by the job execution unit 104. Specifically, the job scheduling unit 101 is a job scheduling program. The job queue 102 is formed in a storage device such as a RAM. Note that the job queue 102 does not hold the job (program) itself, but may hold only information that can identify the job.

  If there is a job in the job queue 102 and the job execution unit 103 has a vacant job execution unit 104 capable of executing a new job, the job scheduling unit 101 determines a predetermined scheduling as described below. According to the method, a job is selected from the job queue 102 and transferred to the job execution means 103. The job execution unit 103 causes the empty job execution unit 104 to execute the new job transferred from the job scheduling unit 101.

  As shown in FIG. 2, the job to be executed includes, in addition to procedure information indicating a procedure necessary for execution, attribute information indicating a parameter at the time of job execution, a job name, and a user name owning the job Meta information indicating the attributes of the job itself is added. The meta information includes a job ID and a job name. The job ID is an identifier for uniquely identifying each job. The job name is an identifier for identifying procedure information. In a general computer system, procedure information corresponds to each step (job step, activity) of a program. The attribute information corresponds to parameters or environment information given to the program. The job ID in the meta information corresponds to a process ID, and the job name corresponds to a program name.

  Next, the operation of the present embodiment will be described with reference to FIG. 1, FIG. 2, and FIG. FIG. 3 is a flowchart illustrating a scheduling method in which the job scheduling unit 101 passes a job waiting for execution in the job queue 102 to the job execution unit 103. When the job execution unit 103 detects that the job execution unit 104 is empty, the job scheduling unit 101 checks whether there is a job waiting for execution in the job queue 103.

  In this example, there are five jobs in the job queue 102, and the job names are “read file”, “read file”, “content comparison”, “content comparison”, “consistency check” in order from the first job. Suppose that The job execution unit 103 has five job execution units 104, four of which are executing jobs, and the job names of the jobs being executed are “file read”, “file read”, and “content comparison”, respectively. ”And“ Consistency Check ”.

  When there is a job waiting for execution, the job scheduling unit 101 selects a job to be passed to the job execution unit 103 according to the procedure shown in FIG. First, the counter N is set to 0 (step S101).

  Next, it is checked whether there are N job names in the job execution unit 103 among the job names of the jobs in the job queue 102 (step S102). Since N = 0 at the beginning, the presence of N means that it does not exist. That is, it is checked whether there is a job name that does not exist. In this example, there is no job with a job name that does not match the job name of any job being executed by the job execution unit 103 in the job queue 102, so the determination result in step S102 is false ("N"). The process proceeds to step S103.

  In step S103, 1 is added to N, and the process returns to step S102. At this stage, N is set to 1 and the process returns to step S102. In step S102, as described above, among job names of jobs in the job queue 102, there are job names that have N job names (one at this stage) among jobs being executed in the job execution unit 103. Check for the presence. In this example, since “content comparison” and “consistency check” are job names of jobs that exist in the job execution unit 103, the determination result in step S102 is true (“Y”), and the process proceeds to step S104. .

  In step S104, a list of job names that exist in the job queue 102 and that exist in the job execution unit 103 (referred to as list A) is created. In this state, “content comparison” and “consistency check” are the contents of list A.

  Next, the job of the job name included in the list A created in step S104 is searched from the head of the job queue 102. If found, the job is taken out from the job queue 102 and transferred to the job execution means 103. In this example, the job having the third “content comparison” job name from the top in the job queue 102 is transferred to the job execution means 103 and executed.

  Here, the case where one job having a job name that matches the job name of the job existing in the job queue 102 exists in the job execution unit 103 is illustrated. If the job execution unit 103 does not have a job having a job name that matches the job name of the job existing in the job queue 102, that is, if N = 0 and the determination result in step S102 is “Y”, the first time With this determination, the process proceeds from step S102 to S104. For example, the names of jobs being executed in the job execution unit 103 are “B”, “C”, “D”, “E”, and the job names of the jobs existing in the job queue 102 are “F” in order. ”,“ G ”, and“ I ”. In this case, a job with the job name “F” is newly submitted to the job execution unit 103 in step S105.

  Further, for example, the names of the jobs executed in the job execution unit 103 are “B”, “B”, “C”, “C”, and the job names of the jobs existing in the job queue 102 are in order. In the case of “B” and “C”, there are two jobs in the job execution unit 103 having a job name that matches the job name of the job existing in the job queue 102. Therefore, in such a case, since N = 2 and the determination result in step S102 is “Y”, the determination in step S102 executed for the third time shifts from step S102 to S104. In this case, a job with the job name “B” is newly submitted to the job execution unit 103 in step S105. This example is the first example.

  Further, for example, the names of jobs executed in the job execution unit 103 are “B”, “B”, “B”, “C”, and the job names of the jobs existing in the job queue 102 are sequentially In the case of “B” and “C”, there is one job in the job execution unit 103 having a job name (“C” in this example) that matches the job name of the job existing in the job queue 102. . Therefore, in such a case, since N = 1 and the determination result in step S102 is “Y”, the determination in step S102 executed for the second time shifts from step S102 to S104. In this case, a job with the job name “C” is newly input to the job execution unit 103 in step S105. This example is a second example.

  Through the processing described above, a job having a job name that is as different as possible from the job name of the job being executed by the job execution unit 103 is newly submitted to the job execution unit 103. More specifically, if the job execution unit 103 does not have a job having a job name that matches the job name of a job that exists in the job queue 102, any job that exists in the job queue 102 (the above example) In this case, the job at the head of the job queue 102) is input to the job execution unit 103, but a job having a job name that matches the job name of the job existing in the job queue 102 exists in the job execution unit 103. If the number of jobs having the same job name (“1” is the smallest) among the jobs existing in the job execution unit 103 is small, “B” in the first example, “B” in the second example, A job having a name that matches “C”) is newly submitted to the job execution means 103.

Embodiment 2. FIG.
Next, a second embodiment of the present invention will be described with reference to the drawings. Referring to FIG. 4, a computer system that implements the job scheduling method according to the second embodiment of the present invention includes a job queue 402 including a plurality of job queues n (n: natural number) holding jobs to be executed, A job scheduling unit 401 for selecting a job to be executed next from jobs existing in the job queue 402, a job execution unit 103 for executing a job, an input / output resource 105 used for job execution, and a computing resource 106 And a memory resource 107.

  A job information table 408 as shown in FIG. In the example shown in FIG. 5, the job information table 408 is set with the number of execution jobs, the number of queue jobs, and the job queue number corresponding to the job name. The number of execution jobs indicates the number of jobs being executed by the job execution unit 103. The queue job number indicates the number of jobs existing in the job queue 402. The job queue number indicates the order of the job queue n in the job queue 402.

  The job execution means 103 includes a plurality of job execution units 104 that actually execute jobs. The job scheduling unit 401 dynamically creates or deletes the job queue n. Jobs to be executed are put in the job queue n. The job scheduling unit 201 updates the job information table 408 by a predetermined method when a job execution request is issued by the system, that is, when a job execution request is issued from a user or a job being executed, for example. Is placed in the job queue n. As described above, the job queue 402 may include a job queue n (n: natural number) corresponding to the job name. Each generated job queue n holds one or more jobs.

  The job scheduling unit 401 also has a job information table according to a predetermined scheduling method if there is a job execution unit 104 that can execute a new job in the job execution unit 103 when there is a job in the job queue 402. 408 is updated. Further, a job is selected from the job queue 402 and transferred to the job execution unit 103. The job execution means 103 causes the empty job execution unit 104 to execute the new job passed from the job scheduling means 401. Further, when the execution of the job is finished, the job scheduling unit 401 is notified of the end of the job by a predetermined method.

  Note that the job to be executed includes attribute information indicating job execution parameters, a job name, and a job in addition to the procedure information indicating the procedure necessary for execution, as in the job in the first embodiment. Meta information indicating the attributes of the job itself such as the user name owned is added.

  Next, the operation of the present embodiment will be described with reference to FIGS. First, the operation of the job scheduling unit 401 when there is a request to queue a job by the system will be described. FIG. 6 is a flowchart showing how the job scheduling unit 401 puts the requested job into the job queue 402. The job scheduling unit 401 refers to the job information table 408 when there is a job queuing request from the system. Then, it is confirmed whether or not there is a line in the job information table 408 where the name N in the job name column matches the job name of the job to be entered in the job queue 402 (step S201). For example, when the job name of the job to be put in the job queue 402 is “file read” (N = “file read”), the job schedule unit 401 sets the value of the job name column in the row of the job information table 408. Check if there is a line that has "read file".

  If there is a line whose value in the job name column is “read file”, the job scheduling unit 401 acquires the job queue number of this line in step S202. When the job information table 408 is in the state as shown in FIG. 5, the job queue number (J) is 3. In addition, the number of job queues in this row is increased by one. In the example shown in FIG. 5, the number of job queues is changed from 2 to 3. In step S205, the job scheduling unit 401 puts the job into the job queue J (job queue 3 in this example) in the job queue 402.

  In step S201, if the job information table 408 does not include a line in the job name column that matches the job name of the job to be entered in the job queue 402, the job scheduling unit 401 sets a new name. A job queue n is created and a unique job queue number is added (step S203). Here, it is assumed that the job queue number of the newly generated job queue n is 3 (J = 3 and n = 3).

  Next, the job scheduling unit 401 adds a new row to the job information table 408 in step S204. The value of each column in the newly added row is the job name that the job name column is to be queued, the job execution number column is 0, the queue job number column is 1, and the job queue number column is new in step S203. This is the job queue number corresponding to the created job queue n. In this example, the job name field is “file read” and the job queue number field is 3. In step S205, the job scheduling unit 401 puts the job to be queued into the job queue n newly created in step S203. Here, the job is placed in the job queue of job queue number 3 in job queue 402, that is, job queue 3.

  Next, the operation of the job scheduling unit 401 when the job execution unit 103 has a free space in the job execution unit 104 and newly executes a job will be described. FIG. 7 is a flowchart of a method by which the job scheduling unit 401 passes a new job to the job execution unit 103.

  When the job scheduling unit 401 passes a new job to the job execution unit 103, the job scheduling unit 401 selects one of the rows of the job information table 408 having the smallest number of job executions (step S301). Let the selected row be row P. For example, when the job information table 408 is formed as shown in FIG. 5, either “Consistency Check” or “Content Comparison” is selected. As described above, when there are a plurality of lines with the smallest number of job executions, it is determined which line is selected by a predetermined determination method. As this determination method, a method using the number of queue jobs, a method using a job queue number, and the like can be considered. Here, it is assumed that a method using a job queue number is adopted and “content comparison” having a small job queue number is selected.

  Next, the job scheduling unit 401 acquires the job queue number (K) of the row P selected in step S301 (step S302). In this example, the value is 1. Further, the number of queue jobs in the bank is reduced by 1, and the number of execution jobs is increased by 1 instead. In this example, the number of queue jobs is 1, and the number of execution jobs is 2. Then, the job scheduling unit 401 takes out one job from the job queue with the job queue number K acquired in step S302, and passes it to the job execution unit 103 (step S303).

  Next, the operation of the job scheduling unit 401 when the job execution unit 103 ends the job will be described. FIG. 8 is a flowchart showing an operation when the job scheduling unit 401 is notified of job completion from the job execution unit 103.

  When the job execution unit 103 is notified of the end of the job, the job scheduling unit 401 acquires the job name M of the completed job from the job execution unit 103 (step S401). The notification of the job name of the completed job may be made together with the notification of job completion. Here, for example, it is assumed that the job having the job name “Consistency Check” is completed (M = “Consistency Check”).

  Next, the job scheduling unit 401 acquires, from the job information table 408, a line in which the value in the job name column matches the job name M acquired in step S401 (step S402). Let the acquired row be row S. If the job information table 408 is formed as shown in FIG. 5, the first line corresponding to the job name “Consistency Check” is acquired. Next, the job scheduling unit 401 decrements the value in the job execution number column in the row of the job information table 408 acquired in step S402 by 1 (step S403). If the job information table 408 is formed as shown in FIG. 5, the value in the job execution column in the first line is changed from 1 to 0.

  Further, the job scheduling unit 401 checks whether both the value in the job execution column and the queue job number column in the row S of the job information table 408 acquired in step S402 are 0 (step S404). If the inspection result in step S404 is true (“Y”), the job scheduling unit 401 deletes the row S of the job information table 408 acquired in step S402 from the job information table and ends the process. If the inspection result in step S404 is false (“N”), the job scheduling unit 401 does nothing and ends the process.

  As described above, also in the present embodiment, in step S301, one of the rows of the job information table 408 divided by job name is selected with the smallest number of job executions and corresponds to the selected row. Since the job to be executed is transferred to the job execution unit 103, a job having a job name different from the job name of the job being executed in the job execution unit 103 is newly input to the job execution unit 103.

  In each of the above embodiments, job scheduling for efficient resource use can be performed without user burden and with low system resource consumption. The reason why efficient resource use can be achieved is that, in general, jobs with the same job name have the same job operation, and the difference is a parameter given to the job, and the nature of resource use tends to be similar. In addition, when each job is part of a large job, each job is specialized for each function, such as input / output part and calculation part. This is because the properties tend to be different. In other words, by distributing the names of jobs to be executed, the resource usage characteristics of jobs are distributed, and resources are used efficiently.

  Also, the reason that there is no burden on the user is that the user does not have to input job resource prediction.

  The reason for the low consumption of system resources is that the system according to each of the above embodiments does not require prior emulation, so that the system resources necessary for scheduling itself can be reduced. Furthermore, the job name used in the scheduling method of each of the above embodiments is information that is always added to the job in a general system, and the burden on the user and system does not increase to use this information. It is.

  The present invention can be effectively applied to a computer system capable of executing a plurality of jobs in parallel.

It is a block diagram which shows the structure of the 1st Embodiment of this invention. It is explanatory drawing which shows the structure of the job in embodiment of this invention. It is a flowchart which shows the operation | movement of the 1st Embodiment of this invention. It is a block diagram which shows the structure of the 2nd Embodiment of this invention. It is explanatory drawing which shows the structural example of the job information table in the 2nd Embodiment of this invention. It is a flowchart which shows the operation | movement of the 2nd Embodiment of this invention. It is a flowchart which shows the operation | movement of the 2nd Embodiment of this invention. It is a flowchart which shows the operation | movement of the 2nd Embodiment of this invention.

Explanation of symbols

101 Job Scheduling Means 102 Job Queue 103 Job Execution Means 104 Job Execution Unit 105 Input / Output Resource 106 Computing Capacity Resource (Computation Resource)
107 Memory Resource 401 Job Scheduling Means 402 Job Queue 408 Job Information Table

Claims (9)

In a job scheduling method for scheduling a job in a computer system capable of executing a plurality of jobs in parallel,
A job scheduling method comprising scheduling according to a job name of a job being executed. The job scheduling method according to claim 1, wherein when a new job becomes executable, a job having a name that matches a name of a job being executed that has a small number of jobs having the same job name is executed. The job scheduling method according to claim 2, wherein a table in which the number of jobs being executed and waiting is set according to a job name is used. In a computer system that can execute multiple jobs in parallel,
Job execution means for executing a plurality of jobs in parallel;
A job queue that holds waiting jobs, and
A computer system comprising: job scheduling means for performing scheduling based on a job name of a job being executed, and transferring the job held in the job queue to the job execution means. When the job execution unit becomes capable of executing a new job, the job scheduling unit selects a job having a name that matches the name of the job that is being executed by the job execution unit with a small number of jobs having the same job name. The computer system according to claim 4, wherein the computer system is taken out from the job queue and passed to the job execution means. Job scheduling means
According to the name of the job, it has a table in which the number of jobs being executed in the job execution means and waiting in the job queue is set,
The job having the name of the job corresponding to the job having the smallest number of jobs in execution in the table is passed to the job execution unit when the job execution unit becomes capable of executing a new job. Computer system. The computer system according to claim 6, wherein the job queue is classified according to a job name. When a job queuing request is made and the job name of the job does not exist in the table, the job scheduling means creates a column corresponding to the job name, and the job execution means notifies the end of the job 8. The computer system according to claim 7, wherein the number of jobs being executed in the column corresponding to the job name of the job is decremented by 1, and the column is deleted when the number of jobs after subtraction becomes 0. A computer having job execution means for executing a plurality of jobs in parallel and a job queue for holding waiting jobs,
When the job execution means becomes capable of executing a new job, a job having a name that matches the name of a job having the same job name with a small number of jobs being executed by the job execution means is designated as the job A job scheduling program for executing processing to be taken out from a queue and delivered to the job execution means.

Images