Automatic preemption in multiple computer systems

• US 8,554,919 B2
• Filed: 09/09/2011
• Issued: 10/08/2013
• Est. Priority Date: 09/09/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Synytskyy, Mykyta, Tamhane, Amitabh Prakash, Dion, David A., Koppolu, Lokesh S
• Primary Examiner(s): Meky, Moustafa M
• Application Number:
  US13/229,415
• Publication Number:
  US 20130067089A1
• Time in Patent Office: 760 Days
• Field of Search: 709200-203, 709217-227, 709/228
• US Class Current: 709/226
• CPC Class Codes: G06F 9/485 Task life-cycle, e.g. stopping, restarting, resuming execution G06F9/4881 takes precedence
• CPC Class Codes: G06F 9/4856 resumption being on a different machine, e.g. task migration, virtual machine migration G06F9/5088 takes precedence

Abstract

Placing an application on a node in a cluster. A method includes determining that a first application cannot be placed on a first node in the first node'"'"'s current running state. The method further includes determining that the first application may be able to be placed on the first node by shutting down or moving one or more other applications running on the first node. The method further includes attempting to shut down or move the one or more other applications on the first node to place the first application.

Claims

1. In a distributed computing environment comprising a cluster, the cluster comprising a plurality of nodes, a method of placing an application on a node in the cluster, the method comprising:

determining that a first application cannot be placed on a first node in the first node'"'"'s current running state;

determining that the first application may be able to be placed on the first node by shutting down or moving one or more other applications running on the first node; and

attempting to shut down or move the one or more other applications on the first node to place the first application, wherein the one or more other application are selected to be shut down based on a determination of a fewest applications affected so as to minimize the number of applications that are shut down to free up resources for the first application.

2. The method of claim 1, further comprising determining that even after attempting to shut down or move the one or more other applications, that the first application cannot be placed on the first node, and as a result selecting a second node on which to place the first application and thereafter repeating acts of claim 1 on the second node.

3. The method of claim 2, further comprising determining that the first application cannot run on any of the nodes in the cluster and ceasing attempting to run the application on a node in the cluster.

4. The method of claim 1, wherein the one or more other application are selected to be shut down based on having a priority that is lower than the first application.

5. The method of claim 1, wherein determining that an application cannot be placed on the first node comprises receiving resource information about the node and applications running on the node.

6. The method of claim 1, wherein determining that the first application may be able to be placed on the first node by shutting down or moving one or more other applications running on the first node is based on received resource information.

7. The method of claim 1 further comprising, attempting to restart one or more of the one or more applications that were shut down.

8. The method of claim 7, wherein determining which application to restart is done in fashion to prevent overloading the cluster by doing an ordered restart.

9. The method of claim 1, wherein processes used to shut down or move the one or more other applications is determined based on relative priorities of the first application and the one or more other applications.

10. The method of claim 1 further comprising, preventing the one or more other applications that were shut down from restarting to prevent a cascade of preemptions and restarts.

11. The method of claim 1, wherein determining that the first application may be able to be placed on the first node by shutting down or moving one or more other applications running on the first node comprises:

determining that the first application has a priority that is greater than the one or more other applications;

determining that the first application has a priority that is greater than or equal to a minimum preemptor priority level; and

determining that the one or more other application have a priority level that is less than or equal to a maximum preemptee priority level.

13. The computer readable medium of claim 12, further comprising determining that even after attempting to shut down or move the one or more other applications, that the first application cannot be placed on the first node, and as a result selecting a second node on which to place the first application and thereafter repeating acts of claim 1 on the second node.

12. In a distributed computing environment comprising a cluster, the cluster comprising a plurality of nodes, a computer readable medium comprising computer executable instructions that when executed by one or more processors, cause the following to be performed:

determining that a first application cannot be placed on a first node in the first node'"'"'s current running state;

determining one or more other applications running on the first node that may be shut down or moved to allow the first application to be placed on the first node; and

attempting to shut down or move the one or more other applications on the first node to place the first application, wherein the one or more other application are selected to be shut down or moved based on a determination of a fewest applications affected so as to minimize the number of applications that are shut down or moved to free up resources for the first application.

14. The computer readable medium of claim 12, wherein the one or more other application are selected to be shut down based on having a priority that is lower than the first application.

15. The computer readable medium of claim 12, wherein determining that an application cannot be placed on the first node comprises receiving resource information about the node and applications running on the node.

16. The computer readable medium of claim 12, wherein determining that the first application may be able to be placed on the first node by shutting down or moving one or more other applications running on the first node is based on received resource information.

17. The computer readable medium of claim 12, further comprising attempting to restart one or more of the one or more applications that were shut down.

18. The computer readable medium of claim 12, wherein processes used to shut down or move the one or more other applications is determined based on relative priorities of the first application and the one or more other applications.

19. The computer readable medium of claim 12 wherein determining that the first application may be able to be placed on the first node by shutting down or moving one or more other applications running on the first node comprises:

determining that the first application has a priority that is greater than the one or more other applications;

determining that the first application has a priority that is greater than or equal to a minimum preemptor priority level; and

determining that the one or more other application have a priority level that is less than or equal to a maximum preemptee priority level.

20. A distributed computing system, the computing system comprising:

a plurality of computers wherein the computer are organized into a computing cluster;

an application hosting module, wherein the application hosting module is configured to;

determine that a first application cannot be placed on a first node in the first node'"'"'s current running state;

determine that the first application can be a preemptor that can preempt other applications on the first node by determining that the application meets a predetermined minimum preemptor priority;

determine that one or more other applications on the first node can be preemptees that can be preempted by the first application to free up enough resources to start the first application; and

determine that the first application may be able to be placed on the first node by shutting down or moving one or more other applications running on the first node, wherein the one or more other application to be shut down or moved are determined based on a determination of a fewest applications affected to minimize the number of applications that are shut down or moved to free up resources for the first application; and

a preemptor module, wherein the preemptor module is configured to attempt to shut down or move the one or more other applications on the first node to place the first application.