Application US20200104184

ACCELERATED RESOURCE ALLOCATION TECHNIQUES
[+] Show Details

US20200104184

1. A method comprising: (7) (5)

determining whether to allocate computer resources for a workload, wherein: the workload is performed for a client device and

the computer resources are part of an edge computing cluster physically proximate to a gateway in communication with the client device;

requesting a recommended recommendation of a computer resource allocation;

applying an artificial intelligence (AI) model to determine the recommendation of the computer resource allocation for the workload, the model to determine the computer resource allocation based at least, in part, on measured performance associated with one or more prior performances of at least a portion of the workload;

determining whether to accept the recommendation of the computer resource allocation; and

in response to a determination to accept the recommendation of the computer resource allocation: causing at least a portion of the edge computing cluster to perform the workload using the recommendation of the computer resource allocation,

determining at least one performance indicator associated with the performance of the workload, and

providing a reward based at least, in part, on the at least one performance indicator.

2. The method of claim 1, wherein the determining whether to allocate computer resources for a workload comprises determining whether to allocate computer resources for a workload based at least in part on one or more of: receipt of one or more of a workload request from a network gateway or client device, an update to hardware or software used by the edge computing resource, a change in service level agreement (SLA) or quality of service (QoS) requirements, or an imminent change in life stage of an application or process. (0)

3. The method of claim 1, wherein the applying an artificial intelligence (AI) model comprises using any or a combination of: a reinforcement learning model, a Q-learning model, a deep-Q learning model, an Asynchronous Advantage Actor-Critic (A3C) model, combinatorial neural network, or recurrent combinatorial neural network. (0)

4. The method of claim 1, wherein the determining whether to accept the recommendation of the computer resource allocation is based at least, in part, on one or more of: use of the recommendation of the computer resource allocation will not yield compliance with a service level agreement (SLA) or the AI model is early in its training phase. (0)

5. The method of claim 1, further comprising: (0) (1)

6. The method of claim 1, wherein providing a reward based at least, in part, on the at least one performance indicator comprises providing a reward based at least, in part, on one or more of: boundedness, end-to-end latency, computer resource utilization, meeting or failing performance requirements, meeting or failing quality of service (QoS) requirements, conflicting requirements of other workloads executing simultaneously on at least portions of the same resources, or rejection of a resource configuration suggestion. (0)

7. The method of claim 1, further comprising: (0) (1)

8. The method of claim 1, further comprising: (0) (5)

receiving a request to perform a second workload;

determining whether at least a portion of the second workload has been performed before using the edge computing cluster;

requesting a recommendation of the second computer resource allocation;

determining whether the second workload is substantially the same as the workload; and

in response to a determination that the second workload is substantially the same as the workload: applying the AI model to determine the recommendation of the second computer resource allocation for the second workload, the AI model to determine the second computer resource allocation based at least, in part, on measured performance associated with the second workload;

causing the edge computing cluster to perform the second workload using the recommendation of the second computer resource allocation;

determining at least one performance indicator associated with the performance of the second workload; and

providing a second reward for the second workload based at least, in part, on the at least one performance indicator, wherein: the at least one performance indicator is based at least, in part, on one or more of: boundedness, end-to-end latency, computer resource utilization, meeting or failing performance requirements, or rejection of a resource configuration suggestion; and

accumulating the second reward with an accumulated award for the workload.

9. An apparatus comprising: (8) (2)

a resource manager for an edge computing cluster and

an interface capable to communicatively couple with the edge computing cluster when connected with the edge computing cluster, wherein the resource manager comprises: at least one processor;

at least one memory communicatively coupled to the at least one processor, wherein the at least one processor is to: identify a resource allocation scenario for a workload A;

apply an artificial intelligence (AI) model to determine a resource allocation recommendation for the workload A, the AI model trained based on rewards arising out of resource allocation recommendations made for one or more workloads that are substantially similar to workload A;

determine whether to accept the resource allocation recommendation for the workload A;

in response to a determination to accept the resource allocation recommendation: cause the edge computing cluster, when coupled to the interface, to use the resource allocation recommendation to perform workload A; and

determine a reward associated with the resource allocation recommendation for workload A to be provided to the AI model.

10. The apparatus of claim 9, wherein to identify a resource allocation scenario for a workload A, the at least one processor is to identify one or more of: receipt of one or more of a workload request from a client device, an update to hardware or software used by the edge computing cluster, a change in service level agreement (SLA) or quality of service (QoS) requirements, or an imminent change in life stage of an application or process. (0)

11. The apparatus of claim 9, wherein the AI model comprises any or a combination of: a reinforcement learning model, a Q-learning model, a deep-Q learning model, or an Asynchronous Advantage Actor-Critic (A3C) model, combinatorial neural network, or recurrent combinatorial neural network. (0)

12. The apparatus of claim 9, comprising a neural network and a graphics processing unit both communicatively coupled to at least one processor, wherein the AI model is to use one or more of: the at least one processor, neural network, and the graphics processing unit. (0)

13. The apparatus of claim 9, wherein to determine whether to accept the resource allocation recommendation for the workload A is based at least, in part, on a determination of one or more of: use of the resource allocation recommendation will not yield compliance with a service level agreement (SLA) or that the AI model is early in its training phase. (0)

14. The apparatus of claim 9, comprising: (1) (1)

a second interface to receive telemetry data, the telemetry data arising out of a performance of the workload A by the edge computing cluster, wherein the at least one processor is to determine at least one of boundedness and end-to-end latency associated with the performance of the workload is based at least, in part, on the telemetry data.

15. The apparatus of claim 14, wherein to determine a reward associated with the resource allocation recommendation for workload A to be provided to the AI model, the at least one processor is to: (0) (1)

16. The apparatus of claim 9, wherein the at least one processor is to: (0) (1)

17. The apparatus of claim 9, further comprising: (0) (1)

18. The apparatus of claim 9, wherein the at least one processor is to: (0) (3)

identify a resource allocation scenario for a workload B;

determine whether workload B is substantially similar to workload A and also is to use at least a portion of the edge computing cluster;

in response to a determination that the workload B is substantially the same as the workload A: apply the AI model to determine a resource allocation recommendation for the workload B;

determine whether to accept the resource allocation recommendation for the workload B;

in response to a determination to accept the resource allocation recommendation: cause the edge computing cluster to use the resource allocation recommendation to perform workload B;

determine a second reward associated with the resource allocation recommendation for workload B to be provided to the AI model, the second reward based at least, in part, on one or more of: boundedness, end-to-end latency, computer resource utilization, meeting or failing performance requirements, or rejection of a resource configuration suggestion; and

store the second reward with an accumulated reward for workload A.

19. A computer-readable medium comprising instructions, that if executed by one or more machines, cause the one or more machines to: (1) (4)

identify a resource allocation scenario for a workload A;

determine whether to accept the resource allocation recommendation for the workload A;

in response to a determination to accept the resource allocation recommendation: cause the edge computing cluster to use the resource allocation recommendation to perform workload A; and

determine a reward associated with the resource allocation recommendation for workload A to be provided to the AI model.

20. The medium of claim 19, wherein to identify a resource allocation scenario for a workload A, the one or more machines are to identify one or more of: receipt of one or more of a workload request from a client device, an update to hardware or software used by the edge computing cluster, a change in service level agreement (SLA) or quality of service (QoS) requirements, or an imminent change in life stage of an application or process. (0)

View Abstract and Specification Size

PDF with Images and Document Face >

Full Text Publication >

Application US20200104184

ACCELERATED RESOURCE ALLOCATION TECHNIQUES

Patent Matrix® Search