GPU resource usage display and dynamic GPU resource allocation in a networked virtualization system

Patent metadata

Field	Value
Publication number	US-10176550-B1
Application number	US-201715463830-A
Country	US
Kind code	B1
Filing date	Mar 20, 2017
Priority date	Mar 20, 2017
Publication date	Jan 8, 2019
Grant date	Jan 8, 2019

Abstract

An architecture for implementing a mechanism for displaying GPU resource usage and dynamically allocating GPU resources in a networked virtualization system is provided. The mechanism compares an initial allocation of GPU resources to virtual machines supported by one or more nodes of the networked virtualization system to a current GPU resource usage by the virtual machines. Based at least in part on the comparison and workloads processed by the virtual machines, the mechanism may reallocate GPU resources to one or more of the virtual machines. A virtual machine is reassigned to a different vGPU profile if reassignment is likely to achieve a more efficient allocation of GPU resources to the virtual machine. A user interface indicating GPU resource usage relative to GPU resource allocation may also be generated and displayed.

First claim

What is claimed is:

1. A method for reallocating resources of a graphics processing unit (GPU) to a virtual machine currently supported on one or more nodes of a networked virtualization system, the method comprising:
retrieving information describing an initial allocation of GPU resources to a plurality of virtual machines, the initial allocation of GPU resources to the plurality of virtual machines corresponding to a first profile associated with a graphics board of one or more graphics boards installed on the one or more nodes;
retrieving information describing a current usage of GPU resources by the plurality of virtual machines;
comparing the initial allocation of GPU resources with the current usage of GPU resources;
determining whether a difference between the initial allocation of GPU resources and the current usage of GPU resources is within a threshold difference; and
responsive to determining the difference between the initial allocation of GPU resources and the current usage of GPU resources is within the threshold difference, for each of the plurality of virtual machines:
retrieving information describing a workload of a virtual machine of the plurality of virtual machines;
comparing the workload of the virtual machine to a workload profile, the workload profile corresponding to a second profile associated with the one or more graphics boards installed on the one or more nodes; and
responsive to determining the workload of the virtual machine has at least a threshold measure of similarity to the workload profile, reallocating GPU resources to the virtual machine, the GPU resources reallocated to the virtual machine corresponding to the second profile associated with the one or more graphics boards installed on the one or more nodes.

2. The method of claim 1 , further comprising:
generating a user interface describing the current usage of GPU resources.

3. The method of claim 2 , wherein the user interface further comprises the difference between the initial allocation of GPU resources and the current usage of GPU resources.

4. The method of claim 2 , wherein the current usage of GPU resources is displayed for each of the plurality of virtual machines.

5. The method of claim 2 , wherein the current usage of GPU resources is displayed graphically.

6. The method of claim 2 , wherein the current usage of GPU resources is expressed as a ratio of virtual machines allocated to the first profile of the one or more graphics boards installed on the one or more nodes.

7. The method of claim 1 , wherein the GPU resources reallocated to the virtual machine comprise an amount of GPU resources that is greater than the initial allocation of GPU resources.

8. The method of claim 1 , wherein the GPU resources reallocated to the virtual machine comprise an amount of GPU resources that is less than the initial allocation of GPU resources.

9. The method of claim 1 , wherein the information describing a current usage of GPU resources by the plurality of virtual machines is retrieved via a set of API calls communicated to the one or more graphics boards.

10. The method of claim 1 , wherein the workload of the virtual machine is compared to a workload profile using a machine-learning model.

11. The method of claim 1 , wherein the information describing the workload of the virtual machine comprises historical workload information.

12. The method of claim 1 , further comprising:
issuing one or more alerts responsive to determining the difference between the initial allocation of GPU resources and the current usage of GPU resources is within the threshold difference.

13. The method of claim 1 , further comprising:
receiving a request from a user to reallocate the GPU resources to the virtual machine.

14. The method of claim 1 , further comprising:
migrating a workload from a first physical GPU associated with the first profile to a second physical GPU associated with the second profile.

15. A computer program product embodied on a non-transitory computer readable medium, the computer readable medium having stored thereon a sequence of instructions which, when executed by a processor causes the processor to execute a method for reallocating resources of a graphics processing unit (GPU) to a virtual machine currently supported on one or more nodes of a networked virtualization system, comprising:
retrieving information describing an initial allocation of GPU resources to a plurality of virtual machines, the initial allocation of GPU resources to the plurality of virtual machines corresponding to a first profile associated with a graphics board of one or more graphics boards installed on the one or more nodes;
retrieving information describing a current usage of GPU resources by the plurality of virtual machines;
comparing the initial allocation of GPU resources with the current usage of GPU resources;
determining whether a difference between the initial allocation of GPU resources and the current usage of GPU resources is within a threshold difference; and
responsive to determining the difference between the initial allocation of GPU resources and the current usage of GPU resources is within the threshold difference, for each of the plurality of virtual machines:
retrieving information describing a workload of a virtual machine of the plurality of virtual machines;
comparing the workload of the virtual machine to a workload profile, the workload profile corresponding to a second profile associated with the one or more graphics boards installed on the one or more nodes; and
responsive to determining the workload of the virtual machine has at least a threshold measure of similarity to the workload profile, reallocating GPU resources to the virtual machine, the GPU resources reallocated to the virtual machine corresponding to the second profile associated with the one or more graphics boards installed on the one or more nodes.

16. The computer program product of claim 15 , wherein the method executed by the processor further comprises:
generating a user interface describing the current usage of GPU resources.

17. The computer program product of claim 16 , wherein the user interface further comprises the difference between the initial allocation of GPU resources and the current usage of GPU resources.

18. The computer program product of claim 16 , wherein the current usage of GPU resources is displayed for each of the plurality of virtual machines.

19. The computer program product of claim 16 , wherein the current usage of GPU resources is displayed graphically.

20. The computer program product of claim 16 , wherein the current usage of GPU resources is expressed as a ratio of virtual machines allocated to the first profile of the one or more graphics boards installed on the one or more nodes.

21. The computer program product of claim 15 , wherein the GPU resources reallocated to the virtual machine comprise an amount of GPU resources that is greater than the initial allocation of GPU resources.

22. The computer program product of claim 15 , wherein the GPU resources reallocated to the virtual machine comprise an amount of GPU resources that is less than the initial allocation of GPU resources.

23. The computer program product of claim 15 , wherein the information describing a current usage of GPU resources by the plurality of virtual machines is retrieved via a set of API calls communicated to the one or more graphics boards.

24. T

Assignees

• Nutanix Inc

Inventors

• Baggerman Cornelis Hendrikus

Classifications

• G06T1/60

  Memory management · CPC title

• G06T1/20 Primary

  Processor architectures; Processor configuration, e.g. pipelining · CPC title

• G06T2200/24

  involving graphical user interfaces [GUIs] · CPC title