MONITORING AND REAL-TIME ADJUSTMENT OF POWER CONSUMPTION SETTINGS

• US 20160224094A1
• Filed: 04/11/2016
• Published: 08/04/2016
• Est. Priority Date: 09/19/2012
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Klein, Matthew David, Marr, Michael David
• Granted Patent: US 9,910,480 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 1/26 Power supply means, e.g. regulation thereof for memories G11C
• CPC Class Codes: G06F 1/28 Supervision thereof, e.g. detecting power-supply failure by out of limits supervision
• CPC Class Codes: G06F 1/3206 Monitoring of events, devices or parameters that trigger a change in power modality
• CPC Class Codes: G06F 1/3234 Power saving characterised by the action undertaken

Abstract

Customers in a multi-tenant environment can obtain energy consumption information for a set of resources or other computing components used by those customers, including time-accurate accounting for various components of those resources utilized on behalf of the customer. A customer can also have the ability to specify how the resources are to be operated when used for the customer, in order to manage the amount of energy consumption. The accounting can be performed even when the resources are shared among multiple users or entities. Various hardware components or agents can be used to provide detailed energy consumption information for those components that is associated with a particular customer. The information can be used not only for accounting and monitoring purposes, but also to make dynamic adjustments based on various changes in usage, energy consumption, or other such factors.

Claims

1. A computer-implemented method of managing energy consumption of resources, comprising:

starting execution of a workload for a customer using at least one virtual machine in a multi-tenant environment, the at least one virtual machine executing on a shared resource within the multi-tenant environment;

receiving an indication that an amount of energy consumption has been utilized by the at least one virtual machine in the execution of the workload in response to an interrupt configured to provide the indication for the amount of energy consumption;

determining a rate of energy consumption, for the execution of the workload using the at least one virtual machine receiving indication of a remaining amount of the workload yet to be executed;

calculating whether using the determined rate of energy consumption for execution of the remaining amount of the workload will satisfy an energy target for the execution of the workload; and

adjusting at least one operational parameter of the at least one virtual machine to adjust the rate of energy consumption when the execution of the remaining amount of the workload is calculated to miss the energy target for execution of the workload.

2. The computer-implemented method of claim 1, further comprising:

monitoring the amount of energy consumed in adjusting the at least one operational parameter and allocating the amount of energy to the execution of the workload.

3. The computer-implemented method of claim 1, wherein the at least one operational parameter includes at least one of an operational mode, an operational state, an operational frequency, or an operational voltage of the at least one virtual machine.

4. The computer-implemented method of claim 1, wherein the interrupt is triggered in response to a threshold being exceeded or an amount of time being reached per a device timer.

5. The computer-implemented method of claim 1, wherein determining a rate of energy consumption includes receiving, from an energy consumption counter, an amount of energy consumed and, from a time stamp counter, an amount of time over which that amount of energy was consumed.

6. The computer-implemented method of claim 1, further comprising:

scheduling execution of the remaining amount of the workload based at least in part upon the determined rate of energy consumption for the execution of the workload.

7. The computer-implemented method of claim 1, wherein calculating whether using the determined rate of energy consumption for execution of the remaining amount of the workload will satisfy an energy target for the execution of the workload includes determining a relative number of different types of instructions included in the remaining amount and calculating based on past energy consumption for workloads with a similar relative number of the different types of instructions.

8. The computer-implemented method of claim 1, further comprising:

receiving information regarding the relative number of different types of instructions from the customer.

9. The computer-implemented method of claim 1, further comprising:

adjusting the at least one operational parameter in response to variations in measured rate of energy consumption over one or more periods of time.

10. The computer-implemented method of claim 1, further comprising:

enabling a user to specify a time target and an energy target for execution of a workload; and

determining a value for each of the at least one operational parameter based at least in part upon the time target and the energy target.

11. A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor of a computing system, cause the computing system to:

receive, in response to an interrupt, notification that an amount of energy consumption has been utilized by at least one virtual machine executing on a shared resource in execution of a workload;

determine a rate of energy consumption for the execution of the workload and a remaining amount of the workload yet to be executed;

calculate whether using the determined rate of energy consumption for the execution of the remaining amount of the workload will satisfy an energy consumption criterion; and

adjust at least one operational parameter of the at least one virtual machine to modify the rate of energy consumption when the execution of the remaining amount of the workload is calculated to not satisfy the energy consumption criterion using current operational parameter values.

12. The non-transitory computer-readable storage medium of claim 11, wherein the notification is received in response to an interrupt for from a user associated with the workload, the interrupt triggered in response to a threshold being exceeded or an amount of time being reached per a device timer.

13. The non-transitory computer-readable storage medium of claim 11, wherein determining the rate of energy consumption includes determining a number of instructions of at least one type that were performed by the at least one virtual machine and determining an amount of energy consumed in performing each instruction of the at least one type.

14. The non-transitory computer-readable storage medium of claim 11, wherein determining a rate of energy consumption includes receiving, from an energy consumption counter, an amount of energy consumed and, from a time stamp counter, an amount of time over which that amount of energy was consumed.

15. The non-transitory computer-readable storage medium of claim 11, wherein execution of the remaining amount of the workload is performed after a period of delay, the period of delay consuming at least one hundred times more energy than adjusting at least one operational parameter.

16. The non-transitory computer-readable storage medium of claim 11, wherein the at least one virtual machine is configured to approximate an amount of energy consumption based at least in part upon a number of instructions flowing through the at least one virtual machine on a specified context.

17. The non-transitory computer-readable storage medium of claim 11, wherein the instructions, when executed by at least one processor of a computing system, further cause the computing system to:

monitor the amount of energy consumed in adjusting the at least one operational parameter and allocating the amount of energy to the execution of the workload.

18. The computer-implemented method of claim 11, wherein the at least one operational parameter includes at least one of an operational mode, an operational state, an operational frequency, or an operational voltage of the at least one virtual machine executing on the shared resource.

19. The non-transitory computer-readable storage medium of claim 11, wherein the instructions, when executed by at least one processor of a computing system, further cause the computing system to:

schedule execution of the remaining amount of the workload based at least in part upon the determined rate of energy consumption for the execution of the workload.

20. The non-transitory computer-readable storage medium of claim 11, wherein determining the rate of energy consumption includes sampling a machine specific register (MSR) that records instantaneous power consumption.