SYSTEM AND METHOD FOR DATA MANAGEMENT JOB PLANNING AND SCHEDULING WITH FINISH TIME GUARANTEE

• US 20090210878A1
• Filed: 02/20/2008
• Published: 08/20/2009
• Est. Priority Date: 02/20/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: HUANG, LAN
• Granted Patent: US 8,230,428 B2
• Time in Patent Office: Days
• US Class Current: 718/102
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues

Abstract

A method is disclosed for scheduling data management jobs on a computer system that uses a dual level scheduling method. Macro level scheduling using a chained timer schedules the data management job for execution in the future. Micro level scheduling using an algorithm controls the actual dispatch of the component requests of a data management job to minimize impact on foreground programs.

Claims

1. A method for use with a data management system, comprising the steps of:

scheduling a data management job using a macro level chained timer to schedule a future data management job;

scheduling the data management job using a micro level virtual clock scheduler to control performing the data management job according to an algorithm; and

executing the data management job according to a finish time generated by the micro level virtual clock scheduler.

2. The method of claim 1, wherein the macro level chained timer is implemented in a management job planner.

3. The method of claim 1, wherein the micro level virtual clock scheduler is implemented in a request scheduler.

4. The method of claim 1, wherein the algorithm comprises the function FT(i,j)=FT(i−

1, j)+1/(φ

j×

C), where FT(i,j) represents finish time of a current component request i of the data management job j, FT(i−

1, j) stands for finish time of a previous component request i−

1 of the data management job j, φ

j stands for a weight of resource reserved for the current data management job, and C stands for overall throughput capacity.

5. The method of claim 4, further comprising the steps of:

coupling a management job planner to a job container populated with data management jobs according to the management job planner;

coupling the job container to the macro level chained timer that timestamps data management jobs and designates a next data management job;

coupling the job container to at least one queue populated with management jobs according to a coupled admission controller;

operating the at least one queue to process data management jobs using a request scheduler to dispatch the data management job from the queue according to the algorithm and considering a feedback input from a resource meter to the request scheduler.

6. The method of claim 5, wherein the resource meter constantly measures an actual resource capacity.

7. The method of claim 5, wherein the request scheduler coupled to the at least one queue receives the data management jobs and dispatches the data management jobs in a proper order to achieve resource sharing.

8. A method for managing data on a computer-based memory to schedule data management jobs, comprising the steps of:

coupling a management job planner to a job container populated with a first group of data management jobs according to the management job planner;

coupling the job container to a timer that timestamps each of the data management jobs to designate a next data management job; and

coupling the job container to a queue populated with a second group of data management jobs including said next data management job, transferred according to a coupled admission controller, said queue of data management jobs processed using a request scheduler to dispatch the next data management job from the queue according to an algorithm and considering a feedback input from a resource meter to the request scheduler.

9. The method of claim 8, further comprising the step of:

measuring the actual resource capacity continually using the resource meter to provide constant feedback input to the request scheduler.

10. The method of claim 8, further comprising the step of:

operating the request scheduler to meet a finish time according to an FT time computed using an algorithm comprising a function FT(i,j)=FT(i−

1, j)+1/(φ

j×

C), where FT(i,j) represents a finish time for a current component request i of the data management job j, FT(i−

1, j) stands for finish time of a previous component request i−

1 of the data management job j, and φ

j stands for a weight of resource reserved for the current data management job, and C stands for overall throughput capacity.

11. The method of claim 8, further comprising the step of:

scheduling a first data management job using a macro level chained timer to schedule the first data management job on the management job planner for future execution.

12. The method of claim 11, further comprising the step of:

scheduling the first data management job using a micro level virtual clock to control performing the first data management job on the request scheduler.

13. The method of claim 12, further comprising the step of:

executing the first data management job according to a finish time generated by the micro level virtual clock on the request scheduler.

14. The method of claim 8, further comprising:

dispatching a plurality of data management jobs from the request scheduler in a proper order according to a macro level chained timer scheduling at the management job planner and micro level virtual clock scheduling at the request scheduler.

15. A method for scheduling data management jobs on a computer system, comprising the steps of:

scheduling a data management job using a macro level chained timer to schedule a macro level time deadline for a future data management job; and

executing component requests of a data management job using a micro level virtual clock to control performing the data management job according to an algorithm FT(i,j)=FT(i−

1, j)+1/(φ

j×

C), where FT(i,j) represents finish time for the of a current component request i of the data management job j, FT(i−

1, j) stands for a finish time of a previous component request i−

1 of the data management job j, φ

j stands for a weight of resource reserved for the current data management job, and C stands for overall throughput capacity.

16. The method of claim 15, further comprising:

populating a data management job container with data management jobs according to a management job planner operating the macro level chained timer;

time stamping data management jobs exiting the job container to designate a next data management job; and

populating a plurality of queues according to a coupled admission controller, with at least one queue data management job processed using a request scheduler to dispatch requests belonging to the at least one data management job from the queue according to the micro level virtual clock algorithm.

17. The method of claim 15, further comprising the step of:

dispatching a request of a data management job from the request scheduler in a proper order according to the macro level scheduling and micro level scheduling.

18. The method of claim 15, wherein the macro scheduling occurs proactively.

19. The method of claim 15, wherein the micro level scheduling occurs reactively.

20. The method of claim 16, further comprising the step of:

controlling the dispatch of at least one queue with an admission controller using a calculated weight of resources reserved, φ

j, based on the macro level time deadline to decide if the system can meet the requirement.