Granted Patent B2

US 12,613,749 · App. 19/096,084 · Granted Apr 28, 2026

Inventors: Zhiming Shen (Sunnyvale, CA); Henrique Fingler (Austin, TX); Robert Wipfel (Draper, UT)

G06F9/5044G06F9/45558G06F9/54G06F2009/45595G06F2209/544

Abstract

Claims (15)

1 . A method for dynamically optimizing and modifying allocation of at least one disaggregated graphical processing unit (GPU) to at least one process via at least one virtual graphical processing units (vGPU), the method comprising:

receiving, by a resource manager executing on a first computing device, from a runtime component executing on a second computing device, a request for allocation, to a process executing on the second computing device, of access to a first GPU in the second computing device;

allocating, by the resource manager, to the process, access to at least one vGPU associated with a second GPU on a third computing device;

transmitting, by the resource manager, to the runtime component, an identification of the at least one vGPU identifying the at least one vGPU as located on the second computing device; and

instantiating, by the runtime component, the process with access to the at least one vGPU.

2 . The method of claim 1 , wherein transmitting, by the resource manager, to the runtime component, the identification of the at least one vGPU includes transmitting an identifier providing information for connecting the first computing device to the second computing device.

3 . The method of claim 1 further comprising:

monitoring, by an agent in communication with the resource manager, utilization of the at least one vGPU by the process;

identifying, by the agent, a modification to a utilization rate by the process of the at least one vGPU;

communicating, by the agent, to the resource manager, the modification;

allocating, by the resource manager, to the process, a second vGPU associated with a third GPU on the third computing device;

and transmitting, by the resource manager, to the runtime component, an identification of the second vGPU identifying the second vGPU as located on the first computing device.

4 . The method of claim 1 further comprising:

determining, by the resource manager, to migrate the vGPU to a fourth computing device; and

transmitting, by the resource manager, to the runtime component, an identification of the vGPU identifying the vGPU as located on the first computing device.

Assignments (3)

Continuity (2)

References Cited (118)