System and method for restarting a workload based on global load balancing

• US 8,984,508 B2
• Filed: 06/02/2011
• Issued: 03/17/2015
• Est. Priority Date: 06/02/2011
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: Shu, Guoqiang, Farkas, Keith, Ma, Eddie, Nelson, Michael, Ziskind, Elisha, Rajagopal, Sridhar, Ji, Minwen
• Primary Examiner(s): Puente, Emerson
• Assistant Examiner(s): Huaracha, Willy W
• Application Number:
  US13/151,862
• Publication Number:
  US 20120311576A1
• Time in Patent Office: 1,384 Days
• Field of Search: 718/1, 718/104, 718/105
• US Class Current: 718/1
• CPC Class Codes: G06F 2009/45575 Starting, stopping, suspending or resuming virtual machine instances
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: G06F 9/4818 Priority circuits therefor
• CPC Class Codes: G06F 9/5077 Logical partitioning of resources; Management or configuration of virtualized resources specific details on emulation or internal functioning of virtual machines G06F9/455

Abstract

A method for restarting a virtual machine in a virtual computing system having a plurality of hosts and a resource scheduler for the plurality of hosts includes writing a placement request for the virtual machine to a shared channel that is accessible by the resource scheduler. The method further includes reading a placement result from the shared channel, wherein the placement result is generated by the resource scheduler responsive to the placement request; and restarting the virtual machine in accordance with the placement result.

Claims

1. A method for restarting a virtual machine in a virtual computing system having a plurality of hosts and a resource scheduler for the plurality of hosts, comprising:

detecting failure of one of the plurality of hosts or a component therein on which the virtual machine depends;

writing a placement request for the virtual machine to a shared channel that is accessible by the resource scheduler;

determining whether a commitment signal is written to the shared channel by the resource scheduler, wherein the commitment signal indicates either that the resource scheduler is able to compute a first placement result within a first amount of time or a second amount of time the resource scheduler is expected to take to compute the first placement result;

if the commitment signal is written to the shared channel and a service objective specifying a guaranteed time during which placement of the virtual machine must be achieved is satisfied, waiting and then reading the first placement result from the shared channel, wherein the first placement result is generated by the resource scheduler responsive to the placement request;

if the commitment signal is not written to the shared channel or the service objective is not satisfied, obtaining a second placement result for the virtual machine from a backup placement engine which makes placement decisions within the guaranteed time specified by the service objective; and

restarting the virtual machine in accordance with the first placement result or the second placement result.

2. The method of claim 1, wherein the first placement result is based on a resource load of at least one of the plurality of hosts.

3. The method of claim 1, further comprising:

in response to not receiving the commitment signal within a threshold amount of time, writing a cancel signal to the shared channel.

4. A method for computing placement results for a virtual machine in a virtual computing system, comprising:

reading a placement request for the virtual machine from a shared channel, the placement request being written to the shared channel upon detecting failure of a host computer or a component therein on which the virtual machine depends;

writing a commitment signal to the shared channel if a first placement result can be computed within an amount of time so as to satisfy a service objective specifying a guaranteed time during which placement of the virtual machine must be achieved, wherein if the first placement result cannot be computed within the amount of time so as to satisfy the service objective, a second placement result is computed by a backup placement engine which makes placement decisions within the guaranteed time specified by the service objective;

if the first placement result can be computed within the amount of time so as to satisfy the service objective, computing the first placement result for the virtual machine based on the placement request read from the shared channel, wherein the first placement result includes information corresponding to a backup host computer on which the virtual machine is to be restarted; and

writing the first placement result to the shared channel, wherein the virtual machine is restarted in accordance with the first placement result or the second placement result.

5. The method of claim 4, wherein a placement engine in a high availability system (HAS) is configured to restart the virtual machine in accordance with the first placement result based on reading the first placement result from the shared channel.

6. The method of claim 4, further comprising:

restarting the virtual machine in accordance with the first placement result; and

writing a status signal to the shared channel indicating that the virtual machine has been restarted.

7. The method of claim 4, wherein computing the first placement result is based on a resource load of at least one of a plurality of hosts configured to execute virtual machines.

8. A non-transitory computer-readable storage medium comprising instructions that, when executed in a computing device, causes the computing device to restart a virtual machine in a virtual computing system having a plurality of hosts and a resource scheduler for the plurality of hosts, by performing the steps of:

detecting failure of one of the plurality of hosts or a component therein on which the virtual machine depends;

writing a placement request for the virtual machine to a shared channel that is accessible by the resource scheduler;

determining whether a commitment signal is written to the shared channel by the resource scheduler, wherein the commitment signal indicates either that the resource scheduler is able to compute a first placement result within a first amount of time or a second amount of time the resource scheduler is expected to take to compute the first placement result;

if the commitment signal is written to the shared channel and a service objective specifying a guaranteed time during which placement of the virtual machine must be achieved is satisfied, waiting and then reading the first placement result from the shared channel, wherein the first placement result is generated by the resource scheduler responsive to the placement request;

if the commitment signal is not written to the shared channel or the service objective is not satisfied, obtaining a second placement result for the virtual machine from a backup placement engine which makes placement decisions within the guaranteed time specified by the service objective; and

restarting the virtual machine in accordance with the first placement result or the second placement result.

9. The non-transitory computer-readable storage medium of claim 8, wherein the first placement result is based on a resource load of at least one of the plurality of hosts.

10. The non-transitory computer-readable storage medium of claim 8, the steps further comprising:

in response to not receiving the commitment signal within a threshold amount of time, writing a cancel signal to the shared channel.

11. A non-transitory computer-readable storage medium comprising instructions that, when executed in a computing device, causes the computing device to compute placement results for a virtual machine in a virtual computing system, by performing the steps of:

reading a placement request for the virtual machine from a shared channel, the placement request being written to the shared channel upon detecting the failure of a host computer or a component therein on which the virtual machine depends;

writing a commitment signal to the shared channel if a first placement result can be computed within an amount of time so as to satisfy a service objective specifying a guaranteed time during which placement of the virtual machine must be achieved, wherein if the first placement result cannot be computed within the amount of time so as to satisfy the service objective, a second placement result is computed by a backup placement engine which makes placement decisions within the guaranteed time specified by the service objective;

if the first placement result can be computed within the amount of time so as to satisfy the service objective, computing the first placement result for the virtual machine based on the placement request read from the shared channel, wherein the first placement result includes information corresponding to a backup host computer on which the virtual machine is to be restarted; and

writing the first placement result to the shared channel, wherein the virtual machine is restarted in accordance with the first placement result or the second placement result.

12. The non-transitory computer-readable storage medium of claim 11, wherein a placement engine in a high availability system (HAS) is configured to restart the virtual machine in accordance with the first placement result based on reading the first placement result from the shared channel.

13. The non-transitory computer-readable storage medium of claim 11, the steps further comprising:

restarting the virtual machine in accordance with the first placement result; and

writing a status signal to the shared channel indicating that the virtual machine has been restarted.

14. The non-transitory computer-readable storage medium of claim 11, wherein computing the first placement result is based on a resource load of at least one of a plurality of hosts configured to execute virtual machines.