Decommissioning physical host machines after relocating unproductive virtual machines therefrom

• US 9,448,841 B2
• Filed: 03/17/2014
• Issued: 09/20/2016
• Est. Priority Date: 03/17/2014
• Status: Active Grant

• Current Assignee: CA, Inc. (d/b/a CA Technologies) (Broadcom, Inc.)
• Original Assignee: CA, Inc. (d/b/a CA Technologies) (Broadcom, Inc.)
• Inventors: Greden, Lara Virginia, Gilbert, Peter James, Mercer, James B., Lankester, Richard John Wcislicki, Ananchaperumal, Dhesikan, Ashby, John Wiley Jr., Pilo, Salvatore, Cattoen, Francois Marie Bruno, Peterson, Paul David
• Primary Examiner(s): Bullock, Jr., Lewis A
• Assistant Examiner(s): Ashley, Bruce S
• Application Number:
  US14/215,995
• Publication Number:
  US 20150261574A1
• Time in Patent Office: 918 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06F 2009/4557 Distribution of virtual machine instances; Migration and load balancing
• CPC Class Codes: G06F 2209/503 Resource availability
• CPC Class Codes: G06F 9/45545 Guest-host, i.e. hypervisor is an application program itself, e.g. VirtualBox
• CPC Class Codes: G06F 9/45558 Hypervisor-specific management and integration aspects
• CPC Class Codes: G06F 9/50 Allocation of resources, e.g. of the central processing unit [CPU]
• CPC Class Codes: G06F 9/5088 involving task migration
• 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 resource management node includes a processor and a memory coupled to the processor. The memory includes computer readable program code that when executed by the processor causes the processor to perform operations. The operations can include identifying an unproductive group of guest virtual machines (VMs) that satisfy a defined rule for being unproductive on an first group of physical host machines based on infrastructure of a distributed computing system used by the first group of the physical host machines to host the guest VMs. A relocation scenario is generated for relocating the unproductive group of the guest VMs from the first group of the physical host machines to a second group of the physical host machines. An effect of the relocation scenario on infrastructure usage of the distributed computing system is estimated. The unproductive group of the guest VMs is relocated from the first group of the physical host machines to the second group of the physical host machines based on the relocation scenario.

Claims

1. A method of operating a resource management node comprising:

identifying an unproductive group of guest virtual machines (VMs) that satisfy a defined rule for being unproductive on a first group of physical host machines based on infrastructure of a distributed computing system used by the first group of the physical host machines to host the guest VMs, wherein the identifying comprises;

for each of a plurality of the guest VMs determining a productivity metric for the guest VM based on infrastructure resources of the distributed computing system that are used by execution of the guest VM on the physical host machine, and controlling a time window over which the productive metric is measured based on an elapsed time needed for the physical host machine to transition from start-up to being ready to execute the guest VM and based on observed change in number of VM guests being executed by the physical host machines over a defined time, including increasing the time window over which the productivity metric is measured responsive to a slower start-up indicated by a longer elapsed time and shortening the time window over which the productivity metric is measured responsive to an increase in the number of VM guests being executed by the physical host machines over the defined time; and

adding to the unproductive group any of the guest VMs having a productivity metric that does not satisfy a defined utilization rule;

generating a relocation scenario for relocating the unproductive group of the guest VMs from the first group of the physical host machines to a second group of the physical host machines to change utilization of the first group of the physical host machines and change utilization of the second group of the physical host machines;

estimating an effect of the relocation scenario on how much infrastructure of the distributed computing system would be used by the first group of the physical host machines and the second group of the physical host machines if the relocation scenario is performed to relocate the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines; and

responsive to the estimate of the effect of the relocation scenario on how much infrastructure of the distributed computing system would be used satisfying a defined rule, relocating the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines by performing the relocation scenario.

2. The method of claim 1,

wherein estimating an effect of the relocation scenario on how much infrastructure of the distributed computing system would be used comprises estimating electrical power consumption of the distributed computing system that will be decreased if the first group of the physical host machines is decommissioned and electrical power consumption by the second group of the physical host machines is increased by processing the unproductive group of the guest VMs;

and

the relocating the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines by performing the relocation scenario, is initiated based on the estimate of the electrical power consumption of the distributed computing system decreasing a threshold amount relative to a present electrical power consumption of the distributed computing system.

3. The method of claim 1, further comprising:

recursively performing the generating a relocation scenario and the estimating an effect of the relocation scenario on how much infrastructure of the distributed computing system would be used until one of relocation scenarios satisfies a defined rule for how much infrastructure of the distributed computing system is allowed to be used.

4. The method of claim 3, further comprising:

relocating the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines by performing the one of the relocation scenarios responsive to determining that the one of the relocation scenarios when completed will satisfy a defined rule for how much electrical power capacity of cabinets, which supply power to the second group of the physical host machines of the distribute computing system, is allowed to be used.

5. The method of claim 3, further comprising:

relocating the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines by performing the one of the relocation scenarios responsive to determining that the one of the relocation scenarios when completed will satisfy a defined rule for how much cooling capacity of cabinets, which cool the second group of the physical host machines of the distributed computer system, is allowed to be used.

6. The method of claim 1, wherein identifying an unproductive group of guest VMs comprises, for each of the guest VMs:

adding the guest VM to the unproductive group if the guest VM does not use at least a threshold amount of resources of the physical host machine to provide the guest VM.

7. The method of claim 1, wherein identifying an unproductive group of guest VMs comprises, for each of the guest VMs:

determining a productivity metric for the guest VM based on resources of the physical host machine that are used to provide the guest VM;

determining, based on the productivity metric for the guest VM, an infrastructure value that indicates how much infrastructure of the distributed computing system is attributed to providing the guest VM; and

adding the guest VM to the unproductive group if the infrastructure value does not satisfy a defined utilization rule for using at least a threshold amount of infrastructure of the physical host machine.

8. The method of claim 1, wherein identifying an unproductive group of guest VMs comprises, for each of the guest VMs:

adding the guest VM to the unproductive group if the guest VM does not satisfy a defined user relevance rule for providing at least a threshold level of client service to users.

9. A resource management node comprising:

a processor; and

a memory coupled to the processor and comprising computer readable program code that when executed by the processor causes the processor to perform operations comprising;

identifying an unproductive group of guest virtual machines (VMs) that satisfy a defined rule for being unproductive on a first group of physical host machines based on infrastructure of a distributed computing system used by the physical host machines to host the guest VMs, wherein the identifying comprises;

for each of a plurality of the guest VMs determining a productivity metric for the guest VM based on infrastructure resources of the distributed computing system that are used by execution of the guest VM on the physical host machine, and controlling a time window over which the productive metric is measured based on an elapsed time needed for the physical host machine to transition from start-up to being ready to execute the guest VM and based on observed change in number of VM guests being executed by the physical host machines over a defined time, including increasing the time window over which the productivity metric is measured responsive to a slower start-up indicated by a longer elapsed time and shortening the time window over which the productivity metric is measured responsive to an increase in the number of VM guests being executed by the physical host machines over the defined time; and

adding to the unproductive group any of the guest VMs having a productivity metric that does not satisfy a defined utilization rule;

generating a relocation scenario for relocating the unproductive group of the guest VMs from the first group of the physical host machines to a second group of physical host machines to change utilization of the first group of the physical host machines and change utilization of the second group of the physical host machines;

estimating an effect of the relocation scenario on how much infrastructure of the distributed computing system would be used by the first group of the physical host machines and the second group of the physical host machines if the relocation scenario is performed to relocate the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines; and

responsive to the estimate of the effect of the relocation scenario on how much infrastructure of the distributed computing system would be used satisfying a defined rule, relocating the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines by performing the relocation scenario.

10. The resource management node of claim 9,

wherein estimating an effect of the relocation scenario on how much infrastructure of the distributed computing system would be used, comprises estimating electrical power consumption of the distributed computing system that will be decreased if the first group of the physical host machines is decommissioned and electrical power consumption by the second group of the physical host machines is increased by processing the unproductive group of the guest VMs;

and

the relocating the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines by performing the relocation scenario, is initiated based on the estimate of the electrical power consumption of the distributed computing system decreasing a threshold amount relative to a present electrical power consumption of the distributed computing system.

11. The resource management node of claim 9, the operations further comprising:

recursively performing the generating a relocation scenario and the estimating an effect of the relocation scenario on how much infrastructure of the distributed computing system would be used until one of relocation scenarios satisfies a defined rule for how much infrastructure of the distributed computing system is allowed to be used.

12. The resource management node of claim 9, wherein the operations further comprise:

relocating the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines by performing the one of the relocation scenarios responsive to determining that the one of the relocation scenarios when completed will satisfy a defined rule for how much electrical power capacity of cabinets, which supply power to the second group of the physical host machines of the distribute computing system, is allowed to be used.

13. The resource management node of claim 9, wherein identifying an unproductive group of guest VMs comprises, for each of the guest VMs:

adding the guest VM to the unproductive group if the guest VM does not satisfy a defined user relevance rule for providing at least a threshold level of client service to users.

14. A computer program product comprising:

a non-transitory computer readable storage medium having computer readable program code embodied in the medium, the computer readable program code comprising;

computer readable code to identify an unproductive group of guest virtual machines (VMs) that satisfy a defined rule for being unproductive on a first group of physical host machines based on infrastructure of a distributed computing system used by the first group of the physical host machines to host the guest VMs,

wherein the identifying comprises;

for each of a plurality of the guest VMs determining a productivity metric for the guest VM based on infrastructure resources of the distributed computing system that are used by execution of the guest VM on the physical host machine, and controlling a time window over which the productive metric is measured based on an elapsed time needed for the physical host machine to transition from start-up to being ready to execute the guest VM and based on observed change in number of VM guests being executed by the physical host machines over a defined time, including increasing the time window over which the productivity metric is measured responsive to a slower start-up indicated by a longer elapsed time and shortening the time window over which the productivity metric is measured responsive to an increase in the number of VM guests being executed by the physical host machines over the defined time; and

adding to the unproductive group any of the guest VMs having a productivity metric that does not satisfy a defined utilization rule;

computer readable code to generate a relocation scenario for relocating the unproductive group of the guest VMs from the first group of the physical host machines to a second group of the physical host machines to change utilization of the first group of the physical host machines and change utilization of the second group of the physical host machines;

computer readable code to estimate an effect of the relocation scenario on how much infrastructure of the distributed computing system would be used by the first group of the physical host machines and the second group of the physical host machines if the relocation scenario is performed to relocate the unproductive group of the guest VMs from the first group of the physical host machines to the second group of the physical host machines; and

computer readable code to, responsive to the estimate of the effect of the relocation scenario on how much infrastructure of the distributed computing system would be used satisfying a defined rule, relocate the unproductive group of the guest VMs from the lint group of the physical host machines to the second group of the physical host machines by performing the relocation scenario.