Energy management for datacenters

• US 8,468,371 B2
• Filed: 12/15/2009
• Issued: 06/18/2013
• Est. Priority Date: 12/15/2009
• Status: Active Grant

• Current Assignee: AirBNB Incorporated
• Original Assignee: International Business Machines Corporation
• Inventors: Krishnamurthy, Rajaram B., Siddique, Naseer S., DeCusatis, Casimer M., Rao, Anuradha, Onghena, Michael
• Primary Examiner(s): Elamin, M
• Application Number:
  US12/637,804
• Publication Number:
  US 20110145603A1
• Time in Patent Office: 1,281 Days
• Field of Search: 713/300, 713/310, 713320-324, 713/330, 713/340
• US Class Current: 713/300
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode

Abstract

A datacenter schedules and executes requests to conserve energy. The datacenter uses an event-based opportunistic approach to schedule and run the requests, which provides energy efficiency. The requests are hierarchically batched and sent to the datacenter for scheduling and execution. They are selectively sent over low power links and selectively serviced by low power processors.

Claims

1. A computer program product for managing requests in a computing environment, the computer program product comprising:

a non-transitory storage medium readable by a processing circuit and storing instructions for execution by the processing circuit for performing a method comprising;

receiving, by at least one processor, a global batch of requests, the global batch of requests comprising a plurality of locally batched requests combined to provide the global batch of requests, the plurality of locally batched requests batched into the global batch of requests based on a selected criterion;

determining whether a request of the global batch of requests is to be scheduled to run, wherein the determining uses a utility function that is based on energy aspects of the computing environment to determine if the request is to be scheduled; and

scheduling the request to run, based on the determining indicating the request is to run.

2. The computer program product of claim 1, wherein the request is to run on a computing unit of a datacenter, said datacenter comprising a plurality of computing units.

3. The computer program product of claim 2, wherein the computing unit to run the request is a low power computing unit.

4. The computer program product of claim 2, wherein the energy aspects include at least one of power cost for the request, utilization of one or more computing units of the datacenter, whether there are one or more computing unit memories of the datacenter in which the request can be piggybacked with one or more other requests, whether there are one or more disks of the datacenter in which the request can be piggybacked, or whether there are one or more power-capped computing units of the datacenter.

5. The computer program product of claim 1, wherein the scheduling includes:

performing at least one of preloading or prelocating data to be used by the request in a queue accessible by one or more computing units to execute the request;

and placing the request on a run queue.

6. The computer program product of claim 1, wherein the method further comprises:

obtaining, by a global dispatcher, the plurality of locally batched requests from one or more computing units coupled to the global dispatcher;

batching, by the global dispatcher, the plurality of locally batched requests to form the global batch of requests; and

forwarding, by the global dispatcher, the global batch of requests to the at least one processor.

7. The computer program product of claim 1, wherein the receiving comprises receiving the global batch of requests using one or more low power links.

8. The computer program product of claim 7, wherein the requests of the global batch of requests are indicated as green requests.

9. The computer program product of claim 1, wherein the request is executing on a computing unit, and the method further comprises:

determining whether the request is to be migrated from the computing unit to another computing unit for energy efficiency.

10. The computer program product of claim 9, wherein the determining comprises deciding whether the computing unit is a low utilization computing unit, and wherein the request is migrated, based on the deciding indicating the computing unit is a low utilization computing unit.

11. The computer program product of claim 1, wherein the method further comprises:

determining whether a request is to be included in a local batch of requests, wherein the determining comprises;

obtaining an energy estimation for the request; and

including the request in the local batch of requests, based on the energy estimation indicating a cost associated with executing the request is acceptable.

21. The computer program product of claim 1, wherein the receiving comprises receiving the global batch of requests from a global dispatcher, the global dispatcher configured to batch the plurality of locally batched requests obtained from one or more computing units coupled to the global dispatcher to provide the global batch of requests.

22. The computer program product of claim 1, wherein the locally batched requests include one or more requests determined to be included in the locally batched requests, wherein a request is determined to be in the locally batched requests based on an energy estimation of the request indicating a cost associated with executing the request is acceptable.

12. A computer system for managing requests in a computing environment, the computer system comprising:

a memory; and

a processor in communications with the memory, wherein the computer system is configured to perform a method, the method comprising;

receiving, by at least one processor, a global batch of requests, the global batch of requests comprising a plurality of locally batched requests combined to provide the global batch of requests, the plurality of locally batched requests batched into the global batch of requests based on a selected criterion;

determining whether a request of the global batch of requests is to be scheduled to run, wherein the determining uses a utility function that is based on energy aspects of the computing environment to determine if the request is to be scheduled; and

scheduling the request to run, based on the determining indicating the request is to run.

13. The computer system of claim 12, wherein the request is to run on a computing unit of a datacenter, said datacenter comprising a plurality of computing units, and wherein the energy aspects include at least one of power cost for the request, utilization of one or more computing units of the datacenter, whether there are one or more computing unit memories of the datacenter in which the request can be piggybacked with one or more other requests, whether there are one or more disks of the datacenter in which the request can be piggybacked, or whether there are one or more power-capped computing units of the datacenter.

14. The computer system of claim 12, wherein the scheduling includes:

performing at least one of preloading or prelocating data to be used by the request in a queue accessible by one or more computing units to execute the request; and

placing the request on a run queue.

15. The computer system of claim 12, wherein the receiving comprises receiving the global batch of requests from a global dispatcher, the global dispatcher configured to batch the plurality of locally batched requests obtained from one or more computing units coupled to the global dispatcher to provide the global batch of requests.

16. The computer system of claim 12, wherein the request is executing on a computing unit, and the method further comprises:

determining whether the request is to be migrated from the computing unit to another computing unit for energy efficiency.

17. The computer system of claim 12, wherein the locally batched requests include one or more requests determined to be included in the locally batched requests, wherein a request is determined to be in the locally batched requests based on an energy estimation of the request indicating a cost associated with executing the request is acceptable.

18. A method of managing requests in a computing environment, said method comprising:

receiving, by at least one processor, a global batch of requests, the global batch of requests comprising a plurality of locally batched requests combined to provide the global batch of requests, the plurality of locally batched requests batched into the global batch of requests based on a selected criterion;

determining whether a request of the global batch of requests is to be scheduled to run, wherein the determining uses a utility function that is based on energy aspects of the computing environment to determine if the request is to be scheduled; and

scheduling, by a processor of the at least one the processor, the request to run, based on the determining indicating the request is to run.

19. The method of claim 18, wherein the request is to run on a computing unit of a datacenter, said datacenter comprising a plurality of computing units, and wherein the energy aspects include at least one of power cost for the request, utilization of one or more computing units of the datacenter, whether there are one or more computing unit memories of the datacenter in which the request can be piggybacked with one or more other requests, whether there are one or more disks of the datacenter in which the request can be piggybacked, or whether there are one or more power-capped computing units of the datacenter.

20. The method of claim 18, wherein the locally batched requests include one or more requests determined to be included in the locally batched requests, wherein a request is determined to be in the locally batched requests based on an energy estimation of the request indicating a cost associated with executing the request is acceptable.