Hybrid acceleration in a processing environment
View Patent ↗Embodiments of the invention include methods and systems for hybrid acceleration in a processing environment. Aspects of the invention include transmitting, by a first computing system to a second computing system, a request for access to an accelerator. The first computing system receives access information for a plurality of accelerators from the second computing system responsive to the request. The first computing system analyzes the access information for the plurality of accelerators to identify a first accelerator from the plurality of accelerators and offloads a first processing job to the first accelerator utilizing the access information.
1. A computer-implemented method for hybrid acceleration, the method comprising:
transmitting, by a first computing system to a second computing system, a request for access to an accelerator, wherein the request for access comprises job requirements for a first processing job;
receiving, by the first computing system, access information for a plurality of accelerators from the second computing system responsive to the request, wherein the access information comprises historic job performance data for each accelerator of the plurality of accelerators and a transfer time for the first processing job to be transferred from the first computing system to the second computing system;
analyzing, by the first computing system, the access information for the plurality of accelerators to determine a first accelerator from the plurality of accelerators that meets the job requirements for the first processing job, wherein determining the first accelerator comprises:
analyzing the historical job performance data associated with the first accelerator and the transfer time to determine a total processing time for the first processing job; and
comparing the total processing time to the job requirements; and
offloading, by the first computing system, the first processing job to the first accelerator utilizing the access information.
2. The computer-implemented method of claim 1 further comprising:
analyzing, by the first computing system, the access information for the plurality of accelerators to identify a second accelerator from the plurality of accelerators; and
offloading a second processing job to the second accelerator utilizing the access information.
3. The computer-implemented method of claim 1 further comprising:
monitoring, by the first computing system, a performance level of the first accelerator.
4. The computer-implemented method of claim 3 further comprising:
based at least in part on a determination that the performance level of the first accelerator is below a threshold performance level, analyzing the access information for the plurality of accelerators to identify a third accelerator from the plurality of accelerators; and
offloading the first processing job to the third accelerator utilizing the access information.
5. The computer-implemented method of claim, 1 wherein the second computing system provides the plurality of accelerators based at least in part on the one or more job requirements.
6. The computer-implemented method of claim 1 , wherein the first computing system is not authorized to access the first accelerator prior to receiving the access information.
7. The computer-implemented method of claim 1 , wherein the first computing system is authorized to access the first accelerator prior to receiving the access information.
8. The computer-implemented method of claim 1 , wherein the first accelerator is located on a separate computing system.
9. The computer-implemented method of claim 1 , wherein the first accelerator is a graphical processing unit.