Selecting and using a cloud-based hardware accelerator

1. An apparatus comprising:

at least one processor;

a memory coupled to the at least one processor; and

an accelerator manager residing in the memory and executed by the at least one processor, the accelerator manager deploying a first accelerator image to a first cloud to generate a first cloud-based hardware accelerator and deploying the first accelerator image to a second cloud to generate a second cloud-based hardware accelerator, wherein the accelerator manager executes a first request on the first cloud-based hardware accelerator, executes the first request on the second cloud-based hardware accelerator, monitors characteristics of the first and second cloud-based hardware accelerators executing the first request, wherein the monitored characteristics comprise execution time for the first and second cloud-based hardware accelerators to process the first request, selects based on the monitored characteristics one of the first and second cloud-based hardware accelerators, and routes subsequent requests to the selected cloud-based hardware accelerator.

2. The apparatus of claim 1 wherein the monitored characteristics further comprise monetary cost of the first and second cloud-based hardware accelerators to process the first request.

3. The apparatus of claim 1 further comprising at least one selection criterion used by the accelerator manager to select the selected cloud-based hardware accelerator.

4. The apparatus of claim 3 wherein the at least one selection criterion comprises execution time for the first and second cloud-based hardware accelerators to process the first request.

5. The apparatus of claim 3 wherein the at least one selection criterion comprises monetary cost of the first and second cloud-based hardware accelerators to process the first request.

6. The apparatus of claim 1 wherein the at least one selection criterion comprises an algorithm based on execution time for the first and second cloud-based hardware accelerators to process the first request and monetary cost of the first and second cloud-based hardware accelerators to process the first request.

7. The apparatus of claim 1 wherein the accelerator manager terminates the first or second cloud-based hardware accelerator that is not the selected cloud-based hardware accelerator.

8. An apparatus comprising:

at least one processor;

a memory coupled to the at least one processor;

a network interface coupled to the at least one processor and coupled to a first cloud and a second cloud; and

an accelerator manager residing in the memory and executed by the at least one processor, the accelerator manager deploying a first accelerator image to the first cloud to generate a first cloud-based hardware accelerator and deploying the first accelerator image to the second cloud to generate a second cloud-based hardware accelerator, wherein the accelerator manager executes a first request on the first cloud-based hardware accelerator, executes the first request on the second cloud-based hardware accelerator, monitors execution time and monetary cost for the first and second cloud-based hardware accelerators to process the first request, selects using an algorithm based on execution time for the first and second cloud-based hardware accelerators to process the first request and monetary cost of the first and second cloud-based hardware accelerators to process the first request.one of the first and second cloud-based hardware accelerators, and routes subsequent requests to the selected cloud-based hardware accelerator.

9. The apparatus of claim 8 wherein the accelerator manager terminates the first or second cloud-based hardware accelerator that is not the selected cloud-based hardware accelerator.

10. A method for using a cloud-based hardware accelerator, the method comprising:

deploying a first accelerator image to a first cloud to generate a first cloud-based hardware accelerator;

deploying the first accelerator image to a second cloud to generate a second cloud-based hardware accelerator;

executing a first request on the first cloud-based hardware accelerator;

executing the first request on the second cloud-based hardware accelerator;

monitoring characteristics of the first and second cloud-based hardware accelerators executing the first request, wherein the monitored characteristics comprise execution time for the first and second cloud-based hardware accelerators to process the first request;

selecting based on the monitored characteristics one of the first and second cloud-based hardware accelerators; and

routing subsequent requests to the selected cloud-based hardware accelerator.

11. The method of claim 10 wherein the monitored characteristics further comprise monetary cost of the first and second cloud-based hardware accelerators to process the first request.

12. The method of claim 10 further comprising at least one selection criterion used by the accelerator manager to select the selected cloud-based hardware accelerator.

13. The method of claim 12 wherein the at least one selection criterion comprises execution time for the first and second cloud-based hardware accelerators to process the first request.

14. The method of claim 12 wherein the at least one selection criterion comprises monetary cost of the first and second cloud-based hardware accelerators to process the first request.

15. The method of claim 10 wherein the at least one selection criterion comprises an algorithm based on execution time for the first and second cloud-based hardware accelerators to process the first request and monetary cost of the first and second cloud-based hardware accelerators to process the first request.

16. The method of claim 10 further comprising terminating the first or second cloud-based hardware accelerator that is not the selected cloud-based hardware accelerator.