Environment based node selection for work scheduling in a parallel computing system

• US 9,122,525 B2
• Filed: 10/30/2013
• Issued: 09/01/2015
• Est. Priority Date: 03/18/2009
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Barsness, Eric L., Darrington, David L., Randles, Amanda, Santosuosso, John M.
• Primary Examiner(s): Cleary, Thomas J
• Application Number:
  US14/067,694
• Publication Number:
  US 20140059556A1
• Time in Patent Office: 671 Days
• Field of Search: 713/300, 713/320, 713/323, 713/324, 718/102, 718/104, 718/105, 702/132, 702/184, 702/185, 702/188, 709/223, 709/224
• US Class Current: 1/1
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode
• CPC Class Codes: G06F 1/329 by task scheduling
• CPC Class Codes: G06F 9/4893 taking into account power or heat criteria power management in computers in general G06F1/3203; thermal management in computers in general G06F1/206
• CPC Class Codes: G06F 9/5094 where the allocation takes into account power or heat criteria power management in computers in general G06F1/3203; thermal management in computers in general G06F1/206
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A method, apparatus, and program product manage scheduling of a plurality of jobs in a parallel computing system of the type that includes a plurality of computing nodes and is disposed in a data center. The plurality of jobs are scheduled for execution on a group of computing nodes from the plurality of computing nodes based on the physical locations of the plurality of computing nodes in the data center. The group of computing nodes is further selected so as to distribute at least one of a heat load and an energy load within the data center. The plurality of jobs may be additionally scheduled based upon an estimated processing requirement for each job of the plurality of jobs.

Claims

1. A method of managing a plurality of jobs throughout a plurality of computing nodes of a parallel computing system disposed in a data center, the method comprising:

in response to receiving a workload, dividing the workload into a plurality of jobs;

accessing a job information table including historical information associated with at least one job of the plurality of jobs, wherein the job information table includes a unique identifier associated with the at least one job;

accessing one or more tables including physical locations associated with each of the plurality of computing nodes and each of one or both of a plurality of power circuits or a plurality of cooling sources distributed throughout the data center, wherein the physical locations are defined by three-dimensional (x,y,z) coordinates stored in the one or more tables;

scheduling the plurality of jobs in the workload for execution on a group of computing nodes from among the plurality of computing nodes in the parallel computing system based upon the physical locations of the plurality of computing nodes in the data center, the physical locations of the one or both of the plurality of power circuits or the plurality of cooling sources, and the historical information in the job information table, including selecting the group of computing nodes by assigning individual computing nodes to form the group so as to distribute at least one of a heat load and an energy load within the data center; and

executing the plurality of jobs on the group of computing nodes.

2. The method of claim 1, wherein the job information table includes historical information indicating an average temperature and an average temperature increase for a core of a computing node associated with processing a particular job, and wherein the plurality of jobs are scheduled such that the heat load is distributed across the data center in a substantially uniform manner based at least in part on at least one of the indicated average temperature and the indicated average temperature increase for a core of a computing node associated with processing a particular job.

3. The method of claim 1, wherein the job information table includes historical information indicating an average energy load and an average increase to an energy load associated with processing a particular job, and wherein the plurality of jobs are scheduled such that the energy load is distributed across the data center in a substantially uniform manner based at least in part on at least one of the indicated average energy load the indicated average increase to an energy load associated with processing a particular job.

4. The method of claim 1, wherein scheduling the plurality of jobs further comprises:

determining physical locations of the plurality of computing nodes in the data center based upon data received from a user interacting with a graphical representation of the data center.

5. The method of claim 1, wherein scheduling the plurality of jobs further comprises:

rescheduling a job from a first computing node to a second computing node.

6. The method of claim 1, wherein scheduling the plurality of jobs further comprises:

determining a computing node heat load for each computing node from among the plurality of computing nodes; and

rescheduling a job from a first computing node with a high computing node heat load to a second computing node with a lower computing node heat load than that of the first computing node.

7. The method of claim 1, wherein scheduling the plurality of jobs further comprises:

determining power circuit energy loads of the plurality of power circuits distributed throughout the data center; and

rescheduling a job from a first computing node that draws energy from a first power circuit with a high power circuit energy load to a second computing node that draws energy from a second power circuit with a lower power circuit energy load than that of the first power circuit.

8. The method of claim 1, wherein the data center is distributed across two or more locations that are geographically distant from each other.

9. The method of claim 1, wherein the data center includes one or more cooling sources, and wherein scheduling the plurality of jobs is additionally based upon the physical locations of the plurality of cooling sources in the data center, including selecting the group of computing nodes so as to distribute the heat load within the data center.

10. The method of claim 9, wherein selecting the group of computing nodes so as to distribute the heat load within the data center includes selecting the group of computing nodes based upon proximity to the plurality of cooling sources.

11. The method of claim 1, wherein scheduling the plurality of jobs is additionally based upon an estimated processing requirement for each job of the plurality of jobs and upon a second indicator separate from a temperature indication for each job, wherein the second indicator characterizes a job based on the job waiting for at least one of data, resource access, and/or for other jobs to complete, the job executing substantially continuously, the job executing with high usage of core processing resources, or the job executing with high usage of processing resources for a long period of time.

12. A method of managing a plurality of jobs throughout a plurality of computing nodes of a parallel computing system disposed in a data center, the method comprising:

in response to receiving a workload, dividing the workload into a plurality of jobs;

accessing a job information table including historical information associated with at least one job of the plurality of jobs, wherein the job information table includes a unique identifier associated with the at least one job;

accessing one or more tables including physical locations associated with each of the plurality of computing nodes and each of one or both of a plurality of power circuits or a plurality of cooling sources distributed throughout the data center, wherein the physical locations are defined by three-dimensional (x,y,z) coordinates stored in the one or more tables;

scheduling the plurality of jobs in the workload for execution on a group of computing nodes from among the plurality of computing nodes in the parallel computing system based upon an estimated processing requirement for each job of the plurality of jobs, the historical information in the job information table, the physical locations of the one or both of the plurality of power circuits or the plurality of cooling sources, and the physical locations of the plurality of computing nodes in the data center , including selecting the group of computing nodes by assigning individual computing nodes to form the group so as to distribute at least one of a heat load and an energy load within the data center; and

executing the plurality of jobs on the group of computing nodes.

13. The method of claim 12, wherein the plurality of jobs are scheduled such that the heat load is distributed across the data center in a substantially uniform manner.

14. The method of claim 12, wherein the plurality of jobs are scheduled such that the energy load is distributed across the data center in a substantially uniform manner.

15. A parallel computing system disposed in a data center, comprising:

a plurality of computing nodes, each computing node including at least one processing unit; and

program code configured to be executed by the parallel computing system to manage a workload of the parallel computing system, the program code further configured to divide the workload into a plurality of jobs, access a job information table including historical information associated with at least one job of the plurality of jobs, wherein the job information table includes a unique identifier associated with at least one job, access one or more tables including physical locations associated with each of the plurality of computing nodes and each of one or both of a plurality of power circuits or a plurality of cooling sources distributed throughout the data center, wherein the physical locations are defined by three-dimensional (x,y,z) coordinates stored in the one or more tables, schedule the plurality of jobs in the workload for execution on a group of computing nodes from among the plurality of computing nodes in the parallel computing system based upon the physical locations of the plurality of computing nodes in the data center, the physical locations of the one or both of the plurality of power circuits or the plurality of cooling sources, and the historical information in the job information table, including selecting the group of computing nodes by assigning individual computing nodes to form the group so as to distribute at least one of a heat load and an energy load within the data center, and execute the plurality of jobs on the group of computing nodes.

16. The system of claim 15, wherein the program code is further configured to schedule the plurality of jobs such that the heat load is distributed across the data center in a substantially uniform manner.

17. The system of claim 15, wherein the program code is further configured to schedule the plurality of jobs such that the energy load is distributed across the data center in a substantially uniform manner.

18. The system of claim 15, wherein the program code is further configured to determine a computing node heat load for each computing node from among the plurality of computing nodes and reschedule a job from a first computing node with a high computing node heat load to a second computing node with a lower computing node heat load than that of the first computing node.

19. The system of claim 15, wherein the program code is further configured to determine power circuit energy loads of the plurality of power circuits and reschedule a job from a first computing node that draws power from a first power circuit with a high power circuit energy load to a second computing node that draws power from a second power circuit with a lower power circuit energy load than that of the first power circuit.

20. The system of claim 15, wherein the program code is further configured to schedule the plurality of jobs based upon the physical locations of the plurality of cooling sources at least one cooling source, including selecting the group of computing nodes so as to distribute the heat load within the data center.