Method of scheduling a job in a clustered computer system and device therefor

• US 5,881,284 A
• Filed: 10/25/1996
• Issued: 03/09/1999
• Est. Priority Date: 10/26/1995
• Status: Expired due to Fees

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Kubo, Hidehito
• Primary Examiner(s): Banankhah, Majid A.
• Application Number:
  US08/738,047
• Time in Patent Office: 865 Days
• Field of Search: 395/671, 395/673, 395/675, 395/672, 707/103
• US Class Current: 718/100
• CPC Class Codes: G06F 9/505 considering the load

Abstract

A computer system and method of scheduling a job in a clustered computer system having a plurality of clusters and a global storage, stores an inputted job in a job queue allocated in the global storage, selects a job to be executed, and executes the selected job in a cluster. The job selection is activated by one of a job completion, a job arrival, and a measurement completion. Upon the job selection, if the resource utilization is low, then a new job is requested. However, if the resource utilization is high, then a new job is not requested.

Claims

1. A computer system having clusters, each of said clusters including at least one processor, said computer system comprising:

a measurement mechanism for measuring a utilization of each cluster of said clusters;

a job submitter for submitting a job to be executed in a cluster of said clusters;

a job controller for controlling the job executed in said cluster of said clusters, and detecting a completion of the job;

a request controller for requesting job selection upon the completion of the job from said job controller according to a measurement result from said measurement mechanism; and

a job selector for selecting a job to be executed in a cluster of said clusters upon one of the job selection request from said request controller and the job submission from said job submitter.

2. The computer system according to claim 1, wherein said request controller further requests the job selection to said job selector upon completion of the measurement by said measurement mechanism.

3. The computer system according to claim 2, wherein said request controller judges whether the utilization of at least one cluster of said clusters has a predetermined status based on said measurement result, and requests said job selector to schedule a job for said at least one cluster of said clusters when the utilization of said at least one cluster is judged to have said predetermined status.

4. The computer system according to claim 3, wherein said request controller upon the completion of a job further judges whether the utilization of a job completed cluster at which the job has just completed has a second predetermined status according to said measurement result, and requests said job selector to schedule a job for said job completed cluster when the utilization of said job completed cluster is judged not to have said second predetermined status.

5. The computer system according to claim 2, wherein said request controller upon the completion of a job judges whether the utilization of a job completed cluster at which the job has just completed has a predetermined status according to said measurement result, and requests said job selector to schedule a job for said job completed cluster when the utilization of said job completed cluster is judged not to have said predetermined status.

6. The computer system according to claim 5, wherein said measurement mechanism measures the utilization of said cluster at every predetermined timing, and said request controller judges that the utilization of said cluster has a second predetermined status, and requests said job selecter to schedule a job for said cluster when the measurement result is continuously less than a first threshold value by a predetermined number of times.

7. The computer system according to claim 5, wherein said measurement mechanism measures the utilization of said cluster at every predetermined timing, said request controller estimates a resource utilization according to a measurement result previously obtained, and said request controller judges that the utilization of said cluster has a second predetermined status, and requests said job selecter to schedule a job for said cluster when the estimated utilization becomes less than a first threshold value.

8. The computer system according to claim 6, wherein said request controller judges that the utilization of at least a first cluster has said predetermined status when a current measurement result of said first cluster exceeds a second threshold value which is higher than the first threshold value.

9. The computer system according to claim 8, wherein said request controller judges that the utilization of at least a second cluster has said second predetermined status when the current measurement result of said second cluster becomes less than a third threshold value which is lower than the first threshold value.

10. The computer system according to claim 7, wherein said request controller judges that the utilization of at least a third cluster has said predetermined status when a current measurement result of said third cluster exceeds a second threshold value which is higher than the first threshold value.

11. The computer system according to claim 1, wherein said job selector is distributed individually for each of said clusters one-by-one.

12. The computer system according to claim 1, wherein only one of said clusters includes said job selector.

13. A method for scheduling a job in a computer system having clusters, each of said clusters including at least one processor, said method comprising steps of:

measuring a utilization of each cluster of said clusters;

submitting a job to be executed in a cluster of said clusters;

detecting a completion of a job being executed;

requesting a first job selection upon the completion of the job according to a result of said measuring step; and

selecting a job to be executed in a cluster of said clusters upon one of the job selection request in said first job selection requesting step and the job submission in said job submitting step.

14. The method for scheduling a job in a computer system according to claim 13, further comprising a step of:

requesting a second job selection upon the completion of a measurement based on a result of said measuring step.

15. The method for scheduling a job in a computer system according to claim 14, wherein:

said requesting a first job selection step includes steps of;

first judging whether the utilization of a job completed cluster of said clusters at which the job has just completed has a first predetermined status according to the result in said measuring step upon the job completion; and

first requesting to schedule a job for said job completed cluster when the utilization of said job completed cluster is judged not to have said first predetermined status in said first judging step, and said requesting a second job selection step includes steps of;

second judging whether the utilization of said at least one cluster has a second predetermined status according to the result in said measuring step upon the completion of a measurement;

second requesting to schedule a job for said at least one cluster of said clusters when the utilization of said at least one cluster is judged to have said second predetermined status in said second judging step.

16. The method for scheduling a job in a computer system according to claim 14, wherein:

said requesting a first job selection step includes steps of;

judging whether the utilization of a job completed cluster has of said clusters at which the job has just completed a predetermined status according to the result in said measuring step upon the job completion; and

requesting to schedule a job for said job completed cluster when the utilization of said at least one cluster is judged not to have said predetermined status in said judging step.

17. The method for scheduling a job in a computer system according to claim 16, wherein said measuring step includes a step of measuring the utilization of said at least one cluster at every predetermined timing, and

wherein said requesting a second job selection step includes a step of judging that the utilization has a second predetermined status when the measurement result is continuously less than a first threshold value by a predetermined number of times.

18. The method for scheduling a job in a computer system according to claim 17, wherein said requesting a first job selection step includes a step of judging that the utilization has said first predetermined status when a current measurement result of said job completed cluster exceeds a second threshold value which is higher than the first threshold value.

19. The method for scheduling a job in a computer system according to claim 18, wherein said requesting a second job selection step includes a step of judging that the utilization has said second predetermined status when the current measurement result of said at least one cluster becomes less than a third threshold value which is lower than the first threshold value.

20. The method for scheduling a job in a computer system according to claim 16, wherein said requesting a second job selection step includes steps of estimating a resource utilization according to a measurement result previously obtained, and judging that the utilization has a second predetermined status when the estimated utilization becomes less than a first threshold value.

21. A computer system having clusters, each of said clusters including at least one processor, said computer system comprising:

a measurement mechanism for measuring a utilization of each cluster of said clusters;

a job submitter for submitting a job to be executed in each cluster of said clusters;

a job controller for controlling the job executed in each cluster of said clusters, and detecting a completion of the job;

a request controller for requesting job selection upon the completion of the job from said job controller according to a measurement result from said measurement mechanism; and

a job selector for selecting a job to be executed in a cluster of said clusters upon one of the job selection request from said request controller and the job submission from said job submitter.