Neural paths for workload execution
View Patent ↗In one example, a neural network model may be generated for a computing network having a plurality of computing resources. Performance metrics and historical usage patterns are the basis for learning current utilization patterns of the computing resources. A workload capacity for each one of the computing resources may be determined, and a neural path may be generated for execution of a specified workload.
1. A computer-readable storage medium including a random access memory (RAM), an electrically-erasable programmable read-only memory (EEPROM), storage drive, or disc having stored thereon instructions which when executed by a processor of a computing device, cause the processor to execute operations comprising:
propagating a neural network model onto a computing network having a plurality of computing resources, wherein the neural network model generates a neural path through the computing network to computing resources providing a workload capacity meeting a workload execution cost associated with executing a workload container, and wherein the workload execution cost is an estimation of computing resources used to execute the workload container or a time for executing the workload container;
sending a report of a respective workload capacity and workload execution cost for each of the plurality of computing resources, wherein the respective workload capacity is based on real-time performance metrics and historical usage patterns of a corresponding computing resource, and wherein the neural path is updated based on changes to the workload capacity of any of the computing resources of the neural path; and
executing the workload container using computing resources of the updated neural path.
2. The computer-readable storage medium of claim 1 , wherein the instructions cause the processor to perform a further operation comprising:
executing the workload container using computing resources of a contingency neural path of the neural network model upon an interruption or disruption of the neural path.
3. The computer-readable storage medium of claim 1 , wherein the real-time performance metrics include CPU utilization, RAM utilization, disk I/O speed, instruction execution speed, network bandwidth and floating-point operation speed.
4. The computer-readable storage medium of claim 1 , wherein the instructions cause the processor to perform a further operation comprising:
storing the neural network model on the computing device.
5. The computer-readable storage medium of claim 1 , wherein the workload execution cost is the time for executing the workload container and wherein the time is a duration of at least one of a simulated workload container execution and an actual workload container execution.
6. The computer-readable storage medium of claim 1 , wherein the instructions cause the processor to perform a further operation comprising:
reporting the execution of the workload container to each of the plurality of computing resources.
7. The computer-readable storage medium of claim 1 , wherein the instructions cause the processor to perform a further operation comprising:
requesting, rejecting, or transferring the workload container based on the workload capacity and the workload execution cost for each computing resource.
8. A method executed by a computing device comprising:
propagating a neural network model onto a computing network having a plurality of computing resources, wherein the neural network model generates a neural path through the computing network to computing resources providing a workload capacity meeting a workload execution cost associated with executing a workload container, and wherein the workload execution cost is an estimation of computing resources used to execute the workload container or a time for executing the workload container;
sending a report of a respective workload capacity and workload execution cost for each of the plurality of computing resources, wherein the respective workload capacity is based on real-time performance metrics and historical usage patterns of a corresponding computing resource, and wherein the neural path is updated based on changes to the workload capacity of any of the computing resources of the neural path; and
executing the workload container using computing resources of the updated neural path.
9. The method of claim 8 , wherein the real-time performance metrics include CPU utilization, RAM utilization, disk I/O speed, instruction execution speed, network bandwidth and floating-point operation speed.
10. The method of claim 8 , further comprising:
executing the workload container using computing resources of a contingency neural path of the neural network model upon an interruption or disruption of the neural path.
11. The method of claim 8 , wherein the workload execution cost is the time for executing the workload container and wherein the time is a duration of at least one of a simulated workload container execution and an actual workload container execution.
12. The method of claim 8 , further comprising:
reporting the execution of the workload container to each of the plurality of computing resources.
13. The method of claim 8 , further comprising:
storing the neural network model on the computing device.
14. The method of claim 8 , further comprising:
requesting, rejecting, or transferring the workload container based on the workload capacity and the workload execution cost for each computing resource.
15. A computing device comprising:
a processor; and
a hardware computer-readable storage medium having stored thereon instructions which when executed by the processor, cause the processor to execute operations comprising:
propagating a neural network model onto a computing network having a plurality of computing resources, wherein the neural network model generates a neural path through the computing network to computing resources providing a workload capacity meeting a workload execution cost associated with executing a workload container, and wherein the workload execution cost is an estimation of computing resources used to execute the workload container or a time for executing the workload container;
sending a report of a respective workload capacity and workload execution cost for each of the plurality of computing resources, wherein the respective workload capacity is based on real-time performance metrics and historical usage patterns of a corresponding computing resource, and wherein the neural path is updated based on changes to the workload capacity of any of the computing resources of the neural path; and
executing the workload container using computing resources of the updated neural path.
16. The computing device of claim 15 , wherein the real-time performance metrics include CPU utilization, RAM utilization, disk I/O speed, instruction execution speed, network bandwidth and floating-point operation speed.
17. The computing device of claim 15 , wherein the instructions cause the processor to perform a further operation comprising:
executing the workload container using computing resources of a contingency neural path of the neural network model upon an interruption or disruption of the neural path.
18. The computing device of claim 17 , wherein the instructions cause the processor to perform a further operation comprising:
storing the neural network model on the computing device.
19. The computing device of claim 15 , wherein the workload execution cost is the time for executing the workload container and wherein the time is a duration of at least one of a simulated workload container execution and an actual workload container execution.
20. The computing device of claim 15 , wherein the instructions cause the processor to perform a further operation comprising:
reporting the execution of the workload container to each of the plurality of computing resources.