Techniques for dynamically assigning jobs to processors in a cluster based on broadcast information

• US 8,122,132 B2
• Filed: 12/16/2008
• Issued: 02/21/2012
• Est. Priority Date: 12/16/2008
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Arimilli, Lakshminarayana Baba, Arimilli, Ravi Kumar, Basso, Claude, Calvignac, Jean L.
• Primary Examiner(s): Coulter, Kenneth R
• Application Number:
  US12/336,312
• Publication Number:
  US 20100153542A1
• Time in Patent Office: 1,162 Days
• Field of Search: 709/226, 709/223, 718/105, 718/104, 718/102, 718/100
• US Class Current: 709/226
• CPC Class Codes: G06F 9/5088 involving task migration

Abstract

A technique for operating a high performance computing cluster (HPC) having multiple nodes (each of which include multiple processors) includes periodically broadcasting information, related to processor utilization and network utilization at each of the multiple nodes, from each of the multiple nodes to remaining ones of the multiple nodes. Respective local job tables maintained in each of the multiple nodes are updated based on the broadcast information. One or more threads are then moved from one or more of the multiple processors to a different one of the multiple processors (based on the broadcast information in the respective local job tables).

Claims

1. A method of operating a high performance computing cluster that includes multiple nodes each including multiple processors, comprising:

periodically broadcasting information, related to processor utilization and network utilization at each of the multiple nodes, from each of the multiple nodes to remaining ones of the multiple nodes;

updating respective local job tables maintained in each of the multiple nodes based on the broadcast information; and

moving, based on the broadcast information in the respective local job tables, one or more threads from one or more of the multiple processors to a different one of the multiple processors.

2. The method of claim 1, further comprising:

moving the one or more threads to a different one of the multiple processors based on communication between the one or more threads.

3. The method of claim 1, further comprising:

moving the one or more threads to a different one of the multiple processors based on workloads of the multiple processors.

4. The method of claim 3, wherein the workloads of the multiple processors are based on utilization of one or more floating-point units included within each of the multiple processors.

5. The method of claim 1, further comprising:

cracking a first thread, included in the one or more threads, that is executing on a first processor, included in the multiple processors, into at least two secondary threads based on a first workload of the first processor; and

moving one of the at least two secondary threads to a second processor, included in the multiple processors, based on a second workload of the second processor.

6. The method of claim 1, wherein the moving further comprises:

moving identified threads, included in the one or more threads, that communicate above a threshold level to identified processors, included in the multiple processors, that are located physically closer to each other.

7. The method of claim 1, wherein the multiple nodes are arranged in a three-dimensional Taurus topology.

8. The method of claim 1, where the multiple processors are included within multiple chip-level multiprocessors.

9. The method of claim 1, wherein at least some of the multiple processors included within different ones of the multiple chip-level multiprocessors are coupled together via host channel adapters and switch fabrics each including one or more switches.

10. The method of claim 1, wherein the information is broadcast using a message passing interface.

11. A high performance computing cluster, comprising:

multiple nodes each including multiple processors, wherein each of the multiple nodes is configured to periodically broadcast information, related to processor utilization and network utilization at each of the multiple nodes, from each of the multiple nodes to remaining ones of the multiple nodes; and

monitoring hardware included in each of the multiple nodes, wherein the monitoring hardware is configured to update respective local job tables maintained in each of the multiple nodes based on the broadcast information, wherein the high performance computing cluster is configured to move, based on the broadcast information in the respective local job tables, one or more threads from one or more of the multiple processors to a different one of the multiple processors.

12. The high performance computing cluster of claim 11, wherein the high performance computing cluster is further configured to move the one or more threads to a different one of the multiple processors based on communication between the one or more threads.

13. The high performance computing cluster of claim 11, wherein the high performance computing cluster is further configured to move the one or more threads to a different one of the multiple processors based on workloads of the multiple processors.

14. The high performance computing cluster of claim 13, wherein the workloads of the multiple processors are based on utilization of one or more floating-point units included within each of the multiple processors.

15. The high performance computing cluster of claim 11, wherein the high performance computing cluster is further configured to crack a first thread, included in the one or more threads, that is executing on a first processor, included in the multiple processors, into at least two secondary threads based on a first workload of the first processor and move one of the at least two secondary threads to a second processor, included in the multiple processors, based on a second workload of the second processor.

16. The high performance computing cluster of claim 11, wherein the high performance computing cluster is further configured to move identified threads, included in the one or more threads, that communicate above a threshold level to identified processors, included in the multiple processors, that are located physically closer to each other.

17. The high performance computing cluster of claim 11, wherein the multiple nodes are arranged in a three-dimensional Taurus topology.

18. The high performance computing cluster of claim 11, where the multiple processors are included within multiple chip-level multiprocessors.

19. The high performance computing cluster of claim 11, wherein at least some of the multiple processors included within different ones of the multiple chip-level multiprocessors are coupled together via host channel adapters and switch fabrics each including one or more switches, and wherein the information is broadcast using a message passing interface.

20. A method of operating a high performance computing cluster that includes multiple nodes each including multiple processors, comprising:

periodically broadcasting information, related to processor utilization and network utilization at each of the multiple nodes, from each of the multiple nodes to remaining ones of the multiple nodes;

updating respective local job tables maintained in each of the multiple nodes based on the broadcast information; and

moving, based on the broadcast information in the respective local job tables, one or more threads from one or more of the multiple processors to a different one of the multiple processors, wherein the information is broadcast using a message passing interface, and wherein at least some of the multiple processors are included within different multiple chip-level multiprocessors that are coupled together via host channel adapters and switch fabrics each including one or more switches.