VGPU scheduling policy-aware migration

1. A system comprising:

at least one computing device comprising at least one processor and at least one data store;

machine readable instructions accessible to the at least one computing device, wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

identify virtual graphics processing unit (vGPU) scheduling migration rules comprising at least one of:

a best-effort matching order that indicates best-effort scheduling, equal-share scheduling, and fixed-share scheduling in decreasing order;

an equal-share matching order that indicates equal-share scheduling, best-effort scheduling, and fixed-share scheduling in decreasing order; and

a fixed-share matching order that indicates fixed-share scheduling, best-effort scheduling, and equal-share scheduling in decreasing order;

identify at least one graphics processing unit (GPU) that is compatible with a current vGPU profile for a virtual machine, wherein the at least one GPU comprises a corresponding at least one vGPU scheduling policy;

determine, based on a current vGPU scheduling policy for the virtual machine, an ordered list of scheduling policies corresponding to: the best-effort matching order, the equal-share matching order, or the fixed-share matching order;

select a destination GPU from the at least one GPU based on a vGPU scheduling policy of the destination GPU being identified as a best available one of the corresponding at least one vGPU scheduling policy according to the ordered list of scheduling policies; and

migrate the virtual machine to the destination GPU.

2. The system of claim 1 , wherein the current vGPU scheduling policy is provided as a key to select the ordered list of scheduling policies.

3. The system of claim 1 , wherein the corresponding at least one vGPU scheduling policy controls how time slices are allocated to vGPUs of the at least one GPU.

4. The system of claim 1 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

configure the destination GPU to a particular vGPU profile based on at least one of: the current vGPU profile, and a GPU memory requirement of the virtual machine.

5. The system of claim 1 , wherein the vGPU scheduling policy corresponds to: best-effort scheduling, equal-share scheduling, or fixed-share scheduling.

6. The system of claim 1 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

identify at least one host that matches threshold migration requirements, wherein the at least one host comprises the at least one GPU.

7. The system of claim 1 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

determine that a number of virtual machines running on the destination GPU is less than a maximum number of virtual machines for the destination GPU.

8. A method performed by at least one computing device executing machine-readable instructions, the method comprising:

Identifying virtual graphics processing unit (vGPU) scheduling migration rules comprising at least one of:

a best-effort matching order that indicates best-effort scheduling, equal-share scheduling, and fixed-share scheduling in decreasing order;

an equal-share matching order that indicates equal-share scheduling, best-effort scheduling, and fixed-share scheduling in decreasing order; and

a fixed-share matching order that indicates fixed-share scheduling, best-effort scheduling, and equal-share scheduling in decreasing order;

identifying at least one graphics processing unit (GPU) that is compatible with a current vGPU profile for a virtual machine, wherein the at least one GPU comprises a corresponding at least one vGPU scheduling policy;

determining, based on a current vGPU scheduling policy for the virtual machine, an ordered list of scheduling policies corresponding to: the best-effort matching order, the equal-share matching order, or the fixed-share matching order;

selecting a destination GPU from the at least one GPU based on a vGPU scheduling policy of the destination GPU being identified as a best available one of the corresponding at least one vGPU scheduling policy according to the ordered list of scheduling policies; and

migrating the virtual machine to the destination GPU.

9. The method of claim 8 , wherein the current vGPU scheduling policy is provided as a key to select the ordered list of scheduling policies.

10. The method of claim 8 , wherein the corresponding at least one vGPU scheduling policy controls how time slices are allocated to vGPUs of the at least one GPU.

11. The method of claim 8 , further comprising:

configuring the destination GPU to a particular vGPU profile based on at least one of: the current vGPU profile, and a GPU memory requirement of the virtual machine.

12. The method of claim 8 , wherein the vGPU scheduling policy corresponds to: best-effort scheduling, equal-share scheduling, or fixed-share scheduling.

13. The method of claim 8 , further comprising:

identifying at least one host that matches threshold migration requirements, wherein the at least one host comprises the at least one GPU.

14. The method of claim 8 , further comprising:

determining that a number of virtual machines running on the destination GPU is less than a maximum number of virtual machines for the destination GPU.

15. A non-transitory computer-readable medium comprising machine readable instructions, wherein the instructions, when executed by at least one processor, cause at least one computing device to at least:

identify virtual graphics processing unit (vGPU) scheduling migration rules comprising at least one of:

a best-effort matching order that indicates best-effort scheduling, equal-share scheduling, and fixed-share scheduling in decreasing order;

an equal-share matching order that indicates equal-share scheduling, best-effort scheduling, and fixed-share scheduling in decreasing order; and

a fixed-share matching order that indicates fixed-share scheduling, best-effort scheduling, and equal-share scheduling in decreasing order;

identify at least one graphics processing unit (GPU) that is compatible with a current vGPU profile for a virtual machine, wherein the at least one GPU comprises a corresponding at least one vGPU scheduling policy;

determine, based on a current vGPU scheduling policy for the virtual machine, an ordered list of scheduling policies corresponding to: the best-effort matching order, the equal-share matching order, or the fixed-share matching order;

select a destination GPU from the at least one GPU based on a vGPU scheduling policy of the destination GPU being identified as a best available one of the corresponding at least one vGPU scheduling policy according to the ordered list of scheduling policies; and

migrate the virtual machine to the destination GPU.

16. The non-transitory computer-readable medium of claim 15 , wherein the current vGPU scheduling policy is provided as a key to select the ordered list of scheduling policies.

17. The non-transitory computer-readable medium of claim 15 , wherein the corresponding at least one vGPU scheduling policy controls how time slices are allocated to vGPUs of the at least one GPU.

18. The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

configure the destination GPU to a particular vGPU profile based on at least one of: the current vGPU profile, and a GPU memory requirement of the virtual machine.

19. The non-transitory computer-readable medium of claim 15 , wherein the vGPU scheduling policy corresponds to: best-effort scheduling, equal-share scheduling, or fixed-share scheduling.

20. The non-transitory computer-readable medium of claim 15 , wherein the instructions, when executed by the at least one processor, cause the at least one computing device to at least:

identify at least one host that matches threshold migration requirements, wherein the at least one host comprises the at least one GPU.