Granted Patent B2

US 12,379,909 · App. 18/308,570 · Granted Aug 5, 2025

Inventors: Dejan S. Milojicic (Milpitas, CA); Aditya Dhakal (Milpitas, CA); Eitan Frachtenberg (Portland, OR); Ninad Sanjay Hogade (Ft. Collins, CO); Rolando Pablo Hong Enriquez (London, GB); Gourav Rattihalli (Milpitas, CA); Tobias Pfandzelter (Milpitas, CA)

Assignee: Hewlett Packard Enterprise Development LP

G06F8/447G06F9/4881G06N5/022G06N20/00

Abstract

Claims (52)

1. A method comprising:

receiving a request to execute a program comprising machine-readable code;

converting, by a system comprising a hardware processor, the machine-readable code into a plurality of functions that are deployable in a Function-as-a-Service (FaaS) computing environment;

determining, by the system, a first deployment of the plurality of functions in the FaaS computing environment, the first deployment defining scheduling and execution tasks for each function at run-time on computing components of the FaaS computing environment;

determining, by a trained prediction model, an execution metric for the first deployment that corresponds with run-time operations of the first deployment;

adjusting, by the system, execution of the plurality of functions and also adjusting the execution metric in a second deployment of the plurality of functions in the FaaS computing environment; and

implementing the second deployment of the plurality of functions in the FaaS computing environment at run-time, the implementing causing execution of the machine-readable code by the computing components of the FaaS computing environment in accordance with the second deployment,

wherein the first deployment and the second deployment are determined prior to run-time.

2. The method of claim 1 , wherein the computing components of the FaaS computing environment comprise one or more of a set including an accelerator, a Field Programmable Gate Array (FPGA), or a Central Processing Unit (CPU).

3. The method of claim 1 , wherein the machine-readable code of the program comprises programmer annotations, and wherein the converting of the machine-readable code into the plurality of functions is based on the programmer annotations.

4. The method of claim 1 , wherein the converting and the determining of the first deployment and the second deployment are performed by a compiler in the system.

5. The method of claim 1 , wherein the computing components of the FaaS computing environment comprise virtual or physical computing components.

6. The method of claim 1 , wherein the converting comprises creating function calls of the plurality of functions in the FaaS computing environment.

7. The method of claim 1 , wherein the determining of the second deployment is part of an iterative adjustment, prior to run-time, of a deployment of the plurality of functions in the FaaS computing environment, the iterative adjustment comprising:

inputting a currently adjusted deployment of the plurality of functions starting with the first deployment to the trained prediction model to produce a predicted execution metric for the currently adjusted deployment that corresponds with run-time operations of the currently adjusted deployment,

determining a modified deployment of the plurality of functions in the FaaS computing environment, the determining of the modified deployment comprising adjusting an assignment of the plurality of functions to the computing components, and

proceeding to a next iteration of the iterative adjustment using the modified deployment as the currently adjusted deployment until the predicted execution metric of the currently adjusted deployment satisfies a threshold.

8. A system comprising:

a processor; and

a non-transitory storage medium storing machine-readable instructions executable on the processor to:

receive a request to execute a program comprising machine-readable code;

convert the machine-readable code into a plurality of functions that are deployable in a Function-as-a-Service (FaaS) computing environment;

determine, prior to run-time of the program, a first deployment of the plurality of functions in the FaaS computing environment, the first deployment defining scheduling and execution tasks for each function at the run-time of the program on computing components of the FaaS computing environment;

determine, by a trained prediction model, an execution metric for the first deployment that corresponds with run-time operations of the first deployment;

determine, prior to the run-time of the program, a second deployment of the plurality of functions in the FaaS computing environment, the determining of the second deployment comprising adjusting an assignment of the plurality of functions to the computing components; and

implement the second deployment of the plurality of functions in the FaaS computing environment at the run-time of the program, the implementing causing execution of the machine-readable code by the computing components of the FaaS computing environment in accordance with the second deployment.

9. The system of claim 8 , wherein the computing components of the FaaS computing environment comprise one or more of an accelerator, an FPGA, or a CPU.

10. The system of claim 8 , wherein the machine-readable code of the program comprises programmer annotations, and the converting of the machine-readable code into the plurality of functions is based on the programmer annotations.

11. The system of claim 8 , wherein the converting and the determining of the first deployment and the second deployment are performed by a compiler in the system.

12. The system of claim 8 , wherein the computing components of the FaaS computing environment comprise virtual or physical computing components.

13. The system of claim 8 , wherein the converting comprises creating function calls of the plurality of functions in the FaaS computing environment.

14. The system of claim 8 , wherein the machine-readable instructions are executable on the processor to:

determine the second deployment as part of an iterative adjustment, prior to the run-time of the program, of a deployment of the plurality of functions in the FaaS computing environment, the iterative adjustment comprising:

inputting a currently adjusted deployment of the plurality of functions starting with the first deployment to the trained prediction model to produce a predicted execution metric for the currently adjusted deployment that corresponds with run-time operations of the currently adjusted deployment,

determining a modified deployment of the plurality of functions in the FaaS computing environment, the determining of the modified deployment comprising adjusting an assignment of the plurality of functions to the computing components, and

proceeding to a next iteration of the iterative adjustment using the modified deployment as the currently adjusted deployment until the predicted execution metric of the currently adjusted deployment satisfies a threshold.

15. A non-transitory computer-readable storage medium comprising instructions that upon execution cause a system to:

receive a request to execute a program comprising machine-readable code;

convert the machine-readable code into a plurality of functions that are deployable in a Function-as-a-Service (FaaS) computing environment;

determine, prior to run-time of the program, a first deployment of the plurality of functions in the FaaS computing environment, the first deployment defining scheduling and execution tasks for each function at the run-time of the program on computing components of the FaaS computing environment;

determine, by a trained prediction model, an execution metric for the first deployment that corresponds with run-time operations of the first deployment;

determine, prior to the run-time of the program, a second deployment of the plurality of functions in the FaaS computing environment, the determining of the second deployment comprising adjusting an assignment of the plurality of functions to the computing components; and

implement the second deployment of the plurality of functions in the FaaS computing environment at the run-time of the program, the implementing causing execution of the machine-readable code by the computing components of the FaaS computing environment in accordance with the second deployment.

16. The non-transitory computer-readable storage medium of claim 15 , wherein the computing components of the FaaS computing environment comprise one or more of an accelerator, an FPGA, or a CPU.

17. The non-transitory computer-readable storage medium of claim 15 , wherein the machine-readable code of the program comprises programmer annotations, and the converting of the machine-readable code into the plurality of functions is based on the programmer annotations.

18. The non-transitory computer-readable storage medium of claim 15 , wherein the converting and the determining of the first deployment and the second deployment are performed by a compiler in the system.

19. The non-transitory computer-readable storage medium of claim 15 , wherein the converting comprises creating function calls of the plurality of functions in the FaaS computing environment.

20. The non-transitory computer-readable storage medium of claim 15 , wherein the instructions upon execution cause the system to:

determine the second deployment as part of an iterative adjustment, prior to the run-time of the program, of a deployment of the plurality of functions in the FaaS computing environment, the iterative adjustment comprising:

inputting a currently adjusted deployment of the plurality of functions starting with the first deployment to the trained prediction model to produce a predicted execution metric for the currently adjusted deployment that corresponds with run-time operations of the currently adjusted deployment,

determining a modified deployment of the plurality of functions in the FaaS computing environment, the determining of the modified deployment comprising adjusting an assignment of the plurality of functions to the computing components, and

proceeding to a next iteration of the iterative adjustment using the modified deployment as the currently adjusted deployment until the predicted execution metric of the currently adjusted deployment satisfies a threshold.

Assignments (3)

Continuity (1)

References Cited (16)