Computing resource forecasting and optimization

• US 10,318,896 B1
• Filed: 09/19/2014
• Issued: 06/11/2019
• Est. Priority Date: 09/19/2014
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Sarkar, Ayan Roy, Ramalingam, Harsha
• Primary Examiner(s): Gart, Matthew S
• Assistant Examiner(s): Holzmacher, Derick J
• Application Number:
  US14/491,762
• Time in Patent Office: 1,726 Days
• Field of Search: None
• CPC Class Codes: G06Q 10/0631 Resource planning, allocation, distributing or scheduling for enterprises or organisations

Abstract

Methods, systems, and computer-readable media for implementing computing resource forecasting and optimization are disclosed. A projected service call volume is determined for a particular service for a future period of time. A set of computing resources is determined to support the projected service call volume for the particular service. The set of computing resources is determined based on automated analysis of the projected service call volume and is calibrated for throughput for the particular service. The set of computing resources is allocated to provide the particular service for the future period of time.

Claims

1. A system, comprising:

a plurality of computing devices configured to implement a computing resource forecasting and optimization system and a service-oriented system, wherein the service-oriented system comprises a plurality of services, and wherein the computing resource forecasting and optimization system is configured to;

determine a projected service call volume for a particular service of the plurality of services for a future period of time;

calibrate one or more computing resources for optimum throughput for the particular service;

determine an optimum set of computing resources to support the projected service call volume for the particular service, wherein the optimum set of computing resources comprises individual computing resources of the computing resources calibrated for optimum throughput for the particular service, wherein the optimum set of computing resources is determined based on a throughput metric;

allocate and configure the optimum set of computing resources to provide the particular service for the future period of time;

determine a second projected service call volume for the particular service for a second future period of time;

determine a second set of computing resources to support the second projected service call volume for the particular service, wherein the second set of computing resources comprises fewer computing resources than the optimum set of computing resources; and

return at least a portion of the optimum set of computing resources to a global pool, wherein the global pool comprises a plurality of computing resources available to one or more other services of the plurality of services.

2. The system as recited in claim 1, wherein at least a portion of the optimum set of computing resources are allocated from a global pool, wherein the global pool comprises a plurality of computing resources available to one or more other services of the plurality of services.

3. The system as recited in claim 1, wherein the computing resource forecasting and optimization system is configured to:

determine that a global pool of computing resources is insufficient to fulfill the optimum set of computing resources, wherein the global pool comprises a plurality of computing resources available to individual services of the plurality of services; and

purchase or lease additional computing resources to fulfill the optimum set of computing resources.

4. The system as recited in claim 1, wherein the computing resource forecasting and optimization system is configured to:

test a plurality of configurations for each computing resource of the one or more computing resources to identify an optimum configuration of each computing resource for optimum throughput.

5. A computer-implemented method, comprising:

determining a projected service call volume for a particular service of a plurality of services for a future period of time;

determining a set of computing resources to support the projected service call volume for the particular service, wherein the set of computing resources is determined based on automated analysis of the projected service call volume, and wherein the set of computing resources is automatically calibrated for throughput for the particular service;

allocating and configuring the set of computing resources to provide the particular service for the future period of time;

determining a second projected service call volume for the particular service for a second future period of time;

determining a second set of computing resources to support the second projected service call volume for the particular service, wherein the second set of computing resources comprises fewer computing resources than the set of computing resources; and

returning at least a portion of the set of computing resources to a global pool, wherein the global pool comprises a plurality of computing resources available to one or more other services of the plurality of services.

6. The method as recited in claim 5, wherein the set of computing resources is determined based on an efficiency index, wherein the efficiency index is determined as an optimum throughput divided by a peak actual throughput for the particular service over a period of time using one or more of the set of computing resources.

7. The method as recited in claim 5, further comprising:

determining a forecast for one or more demand drivers for the future period of time, wherein the projected service call volume is determined based on the forecast for the one or more demand drivers.

8. The method as recited in claim 5, wherein the projected service call volume is determined based on automated analysis of trace data for individual services of the plurality of services, wherein the automated analysis of the trace data determines one or more relationships between the particular service and individual demand drivers.

9. The method as recited in claim 5, wherein the projected service call volume is determined based on a time-series analysis of service call volume history for the particular service.

10. The method as recited in claim 5, wherein at least a portion of the set of computing resources are allocated from a global pool, wherein the global pool comprises a plurality of computing resources available to one or more other services of the plurality of services.

11. The method as recited in claim 5, further comprising:

determining that a global pool of computing resources is insufficient to fulfill the set of computing resources, wherein the global pool comprises a plurality of computing resources available to individual services of the plurality of services; and

purchasing or leasing additional computing resources to fulfill the set of computing resources.

12. The method as recited in claim 5, further comprising:

testing a plurality of configurations for each computing resource of the one or more computing resources to identify an optimum configuration of each computing resource for optimum throughput.

13. A non-transitory, computer-readable storage medium storing program instructions computer-executable to perform:

determining a projected service call volume for a particular service of a plurality of services for a future period of time;

determining a set of computing resources to support the projected service call volume for the particular service, wherein the set of computing resources is determined based on automated analysis of the projected service call volume, and wherein the set of computing resources is automatically calibrated for throughput for the particular service;

allocating and configuring the set of computing resources to provide the particular service for the future period of time;

determining a second projected service call volume for the particular service for a second future period of time;

determining a second set of computing resources to support the second projected service call volume for the particular service, wherein the second set of computing resources comprises fewer computing resources than the set of computing resources; and

returning at least a portion of the set of computing resources to a global pool, wherein the global pool comprises a plurality of computing resources available to one or more other services of the plurality of services.

14. The non-transitory, computer-readable storage medium as recited in claim 13, wherein the set of computing resources is determined based on an efficiency index, wherein the efficiency index is determined as an optimum throughput divided by a peak actual throughput for the particular service over a period of time using one or more of the set of computing resources.

15. The non-transitory, computer-readable storage medium as recited in claim 13, wherein the program instructions are further computer-executable to perform:

determining a forecast for one or more demand drivers for the future period of time, wherein the projected service call volume is determined based on the forecast for the one or more demand drivers.

16. The non-transitory, computer-readable storage medium as recited in claim 13, wherein the projected service call volume is determined based on automated analysis of trace data for individual services of the plurality of services, wherein the automated analysis of the trace data determines one or more relationships between the particular service and individual demand drivers.

17. The non-transitory, computer-readable storage medium as recited in claim 13, wherein the projected service call volume is determined based on a time-series analysis of service call volume history for the particular service.

18. The non-transitory, computer-readable storage medium as recited in claim 13, wherein at least a portion of the set of computing resources are allocated from a global pool, wherein the global pool comprises a plurality of computing resources available to one or more other services of the plurality of services.

19. The non-transitory, computer-readable storage medium as recited in claim 13, wherein the program instructions are further computer-executable to perform:

determining that a global pool of computing resources is insufficient to fulfill the set of computing resources, wherein the global pool comprises a plurality of computing resources available to individual services of the plurality of services; and

purchasing or leasing additional computing resources to fulfill the set of computing resources.

20. The non-transitory, computer-readable storage medium as recited in claim 13, wherein the program instructions are further computer-executable to perform:

monitoring performance of the set of computing resources after the set of resources is calibrated for throughput for the particular service;

determining that the performance is not within predetermined boundaries; and

in response to determining that the performance is not within the predetermined boundaries, recalibrating the set of computing resources for throughput for the particular service.