VIRTUAL MACHINE POWER CONSUMPTION MEASUREMENT AND MANAGEMENT

• US 20110213997A1
• Filed: 02/26/2010
• Published: 09/01/2011
• Est. Priority Date: 02/26/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Bhattacharya, Arka Aloke, Liu, Jie, Kansal, Aman, Burger, Douglas C.
• Granted Patent: US 8,862,914 B2
• Time in Patent Office: Days
• US Class Current: 713/324
• CPC Class Codes: G01R 21/133 by using digital technique
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode
• CPC Class Codes: G06F 1/3206 Monitoring of events, devices or parameters that trigger a change in power modality
• CPC Class Codes: G06F 1/3228 Monitoring task completion, e.g. by use of idle timers, stop commands or wait commands
• CPC Class Codes: G06F 1/3234 Power saving characterised by the action undertaken
• CPC Class Codes: G06F 1/329 by task scheduling
• CPC Class Codes: G06F 11/3409 for performance assessment
• CPC Class Codes: G06F 11/3423 where the assessed time is active or idle time
• CPC Class Codes: G06F 11/3447 Performance evaluation by modeling
• CPC Class Codes: G06F 11/3466 Performance evaluation by tracing or monitoring
• CPC Class Codes: G06F 2201/815 Virtual middleware or OS functionality using virtual machines to implement generic software techniques for error detection or fault masking G06F11/1484
• CPC Class Codes: G06F 2201/865 Monitoring of software
• CPC Class Codes: G06F 2201/88 Monitoring involving counting
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

Embodiments of the virtual machine power metering system and method measure the power consumption of individual virtual machines. Power meter measurements for a physical host server are converted into individual virtual machine power meters that measure the power consumption of each individual virtual machine residing on the host server. The virtual machine power consumption is computed by generating a power model using the total power consumption of the host server and resource utilization for a virtual machine. Optimal power model coefficients are computed using the power model. The energy used by the virtual machine is computed using one of two embodiments. Embodiments of the system and method also can be used to obtain the power consumption for a specific activity (such as a service, request, or search query). In addition, the virtual machine power metering can be used for virtual machine power capping to allow power oversubscription in virtualized environments.

Claims

1. A method implemented on a computing device for computing power consumption of an activity that occurs across at least one virtual machine residing on a physical host server, comprising:

defining an activity period as an amount of time elapsed between a beginning event and an ending event;

tracking each resource on each virtual machine that is used in the activity during the activity period;

computing power consumption for each virtual machine working on the activity during the activity period to obtain power consumption measurements;

summing the power consumption measurements for each virtual machine working on the activity during the activity period to obtain the total activity power consumption; and

outputting the total activity power consumption representing the total power consumed for the activity.

2. The method of claim 1, further comprising:

computing a power model that relates a total resource utilization for a virtual machine to a total power consumption of the host server; and

computing a power consumption for the virtual machine using the power model and optimal power model coefficients.

3. The method of claim 1, further comprising:

computing total power consumption of devices used during the activity period but not tracked as a resource; and

determining a portion of the total power consumption of the untracked devices that was used in the activity during the activity period.

4. The method of claim 2, further comprising summing the power consumption measurements for each virtual machine working on the activity and the portion of the power consumption of the untracked devices using during the activity period to obtain the total activity power consumption.

5. The method of claim 1, further comprising defining an activity as at least one of:

(a) a service;

(b) a request;

(c) a search query;

(d) an e-mail delivery;

(e) a webpage retrieval;

(f) a database operation;

(g) a file update.

6. The method of claim 4, further comprising:

generating a beginning event when the activity is initiated on the virtual machine; and

determining and marking an ending event when the activity is terminated and presented to a user.

7. The method of claim 5, further comprising reporting the total activity power consumption as:

(a) a single number in units of energy per virtual machine activity;

(b) a single number in units of energy per virtual machine over the activity period;

or (c) a time series with multiple values in units of power over the activity period.

8. The method of claim 2, further comprising:

determining percentages of each resource used by each virtual machine;

multiplying each resource percentage by its corresponding optimal power model coefficients to obtain intermediate resource results for each virtual machine; and

summing each of the intermediate resource results for each virtual machine to obtain the individual power consumption for each virtual machine.

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

determining percentages of a central processing unit (CPU) working for each virtual machine; and

using the optimal power model coefficients to compute virtual machine-specific coefficients for each resource for each virtual machine.

10. A method implemented on a computing device for individually controlling power consumption of a plurality of virtual machines residing on a physical host server, comprising:

using the computing device to perform the following;

monitoring a power consumption of each virtual machine on the host server;

measuring a peak power drawn by each virtual machine on the host server;

determining whether any of the plurality of virtual machines has exceeded its power budget;

if so, then determining whether a power budget of the host server has been exceeded; and

if so, then reducing a power consumption of at least one of the virtual machines to reduced a total power consumption of the host server.

11. The method of claim 10, further comprising locating virtual machines on the host server based on measured peak power of each virtual machine.

12. The method of claim 10, further comprising:

generating a power model that relates a total resource utilization for each of the plurality of virtual machines to a total power consumption of the host server;

using the power model to obtain optimal power model coefficients; and

computing a power consumption for each of the plurality of virtual machines using the power model and the optimal power model coefficients.

13. The method of claim 10, further comprising reducing power consumption of each virtual machine that is exceeding its power budget so as to not exceed the power budget of the host server.

14. The method of claim 10, further comprising reducing power consumption of each lower-priority virtual machines residing the host server so as to not exceed the power budget of the host server, where the lower-priority virtual machines have a lower priority as compared to virtual machines that have exceeded their power budget.

15. The method of claim 10, further comprising shutting down one or more of the virtual machines that have exceeded their power budget so as to not exceed the power budget of the host server.

16. The method of claim 10, further comprising migrating one or more virtual machines to another server so as to not exceed the power budget of the host server.

17. A computer-implemented method for generating summary images for computing an individual power consumption of a virtual machine residing on a physical host server, comprising:

obtaining a total power consumption for the host server;

determining a resource utilization for each resource on the host server to find out how much of each resource the virtual machine is using;

summing each resource utilization for each resource to obtain a total resource utilization;

generating a power model that relates the total resource utilization for the virtual machine to the total power consumption of the host server;

determining M number of resources that are being used by the virtual machine;

making N number of measurements of the total power consumption for N different values of resource utilization to obtain a system of N equations and M unknowns;

solving the system of N equations to obtain M number of optimal power model coefficients; and

computing the individual power consumption of the virtual machine using the power model and the optimal power model coefficients.

18. The computer-implemented method of claim 17, further comprising:

determining percentages of each resource used by the virtual machine;

multiplying each resource percentage by its corresponding optimal power model coefficients to obtain intermediate resource results; and

summing each of the intermediate resource results to obtain the individual power consumption for the virtual machine;

19. The computer-implemented method of claim 17, further comprising:

determining percentages of a central processing unit (CPU) working for the virtual machine; and

computing virtual machine-specific coefficients for each resource using the optimal power model coefficients.

20. The computer-implemented method of claim 19, further comprising:

multiplying each CPU percentage and resource percentage by its corresponding virtual machine-specific coefficients to obtain intermediate resource results; and

summing each of the intermediate resource results to obtain the individual power consumption for the virtual machine.