Energy management of a computing deployment with guaranteed end-to-end performance

Patent metadata

Field	Value
Publication number	US-2016170476-A1
Application number	US-201414567939-A
Country	US
Kind code	A1
Filing date	Dec 11, 2014
Priority date	Dec 11, 2014
Publication date	Jun 16, 2016
Grant date	—

Abstract

Techniques for managing energy use of a computing deployment are provided. In one embodiment, a computer system can establish a performance model for one or more components of the computing deployment, where the performance model models a relationship between one or more tunable parameters of the one or more components and an end-to-end performance metric, and where the end-to-end performance metric reflects user-observable performance of a service provided by the computing deployment. The computer system can further execute an algorithm to determine values for the one or more tunable parameters that minimize power consumption of the one or more components, where the algorithm guarantees that the determined values will not cause the end-to-end performance metric, as calculated by the performance model, to cross a predefined threshold. The computer system can then enforce the determined values by applying changes to the one or more components.

First claim

What is claimed is:

1 . A method for managing energy use of a computing deployment, the method comprising:
establishing, by a computer system, a performance model for one or more components of the computing deployment, the performance model being configured to model a relationship between one or more tunable parameters of the one or more components and an end-to-end performance metric, the end-to-end performance metric reflecting user-observable performance of a service provided by the computing deployment; executing, by the computer system, an algorithm to determine values for the one or more tunable parameters that minimize power consumption of the one or more components, the algorithm guaranteeing that the determined values will not cause the end-to-end performance metric, as calculated by the performance model, to cross a predefined threshold; and enforcing, by the computer system, the determined values by applying changes to the one or more components.

2 . The method of claim 1 wherein establishing the performance model comprises:
adjusting values for the one or more tunable parameters over allowable ranges;
for each combination of adjusted values, measuring a value of the end-to-end performance metric; and
generating the performance model based on the adjusted values and the measured values.

3 . The method of claim 1 wherein the computing deployment includes a plurality of servers configured to host virtual machines (VMs), and wherein the algorithm determines values for the one or more tunable parameters that minimize power consumption of the plurality of servers.

4 . The method of claim 3 wherein the one or more tunable parameters include a CPU frequency and a VM consolidation ratio for each of the plurality of servers.

5 . The method of claim 4 wherein executing the algorithm comprises:
determining a VM consolidation ratio that can be supported by each server in the plurality of servers in view of the performance model and the predefined threshold, assuming the server's CPU runs at maximum frequency;
calculating, based on the determined VM consolidation ratio and a number of VM users, a total number of active servers needed; and
if the performance model indicates that, for each server, the end-to-end performance metric does not cross the predefined threshold when using the determined VM consolidation ratio and the maximum CPU frequency, calculating a CPU frequency that minimizes a power consumption function for the server, subject to a constraint that the CPU frequency cannot cause the end-to-end performance metric to cross the predefined threshold.

6 . The method of claim 5 wherein enforcing the determined values comprises:
migrating VMs and shutting down idle servers based on the total number of active servers needed; and
for each active server, throttling the server's CPU based on the calculated CPU frequency.

7 . The method of claim 3 wherein the computing deployment further includes storage components and networking components, and wherein the algorithm determines values for the one or more tunable parameters that further minimize power consumption of the storage components and networking components.

8 . The method of claim 1 wherein the service is a virtual desktop infrastructure (VDI) service.

9 . A non-transitory computer readable storage medium having stored thereon program code executable by a computer system, the program code embodying a method for managing energy use of a computing deployment, the method comprising:
establishing a performance model for one or more components of the computing deployment, the performance model being configured to model a relationship between one or more tunable parameters of the one or more components and an end-to-end performance metric, the end-to-end performance metric reflecting user-observable performance of a service provided by the computing deployment; executing an algorithm to determine values for the one or more tunable parameters that minimize power consumption of the one or more components, the algorithm guaranteeing that the determined values will not cause the end-to-end performance metric, as calculated by the performance model, to cross a predefined threshold; and enforcing the determined values by applying changes to the one or more components.

10 . The non-transitory computer readable storage medium of claim 9 wherein establishing the performance model comprises:
adjusting values for the one or more tunable parameters over allowable ranges;
for each combination of adjusted values, measuring a value of the end-to-end performance metric; and
generating the performance model based on the adjusted values and the measured values.

11 . The non-transitory computer readable storage medium of claim 9 wherein the computing deployment includes a plurality of servers configured to host virtual machines (VMs), and wherein the algorithm determines values for the one or more tunable parameters that minimize power consumption of the plurality of servers.

12 . The non-transitory computer readable storage medium of claim 11 wherein the one or more tunable parameters include a CPU frequency and a VM consolidation ratio for each of the plurality of servers.

13 . The non-transitory computer readable storage medium of claim 12 wherein executing the algorithm comprises:
determining a VM consolidation ratio that can be supported by each server in the plurality of servers in view of the performance model and the predefined threshold, assuming the server's CPU runs at maximum frequency;
calculating, based on the determined VM consolidation ratio and a number of VM users, a total number of active servers needed; and
if the performance model indicates that, for each server, the end-to-end performance metric does not cross the predefined threshold when using the determined VM consolidation ratio and the maximum CPU frequency, calculating a CPU frequency that minimizes a power consumption function for the server, subject to a constraint that the CPU frequency cannot cause the end-to-end performance metric to cross the predefined threshold.

14 . The non-transitory computer readable storage medium of claim 13 wherein enforcing the determined values comprises:
migrating VMs and shutting down idle servers based on the total number of active servers needed; and
for each active server, throttling the server's CPU based on the calculated CPU frequency.

15 . The non-transitory computer readable storage medium of claim 11 wherein the computing deployment further includes storage components and networking components, and wherein the algorithm determines values for the one or more tunable parameters that further minimize power consumption of the storage components and networking components.

16 . The non-transitory computer readable storage medium of claim 9 wherein the service is a virtual desktop infrastructure (VDI) service.

17 . A computer system comprising:
a processor; a performance modeler component executed by the processor, the performance modeler component being configured to establish a performance model for one or more components of a computing deployment, the performance model being configured to model a relationship between one or more tunable parameters of the one or more components and an end-to-end performance metric, the end-to-end performance metric reflecting user-observable performance of a service provided by the computing deployment; an optimizer component executed by the processor, the optimizer component being configured to execute an algorithm to dete

Assignees

• Vmware Inc

Inventors

• Fu Xing
• Magdon-Ismail Tariq

Classifications

• G06F11/3062

  where the monitored property is the power consumption (power management in a computing system G06F1/3203) · CPC title

• H04L41/0813

  characterised by the conditions triggering a change of settings · CPC title

• G06F2201/81

  Threshold · CPC title

• H04L41/0833

  for reduction of network energy consumption · CPC title

• G06F11/3409

  for performance assessment · CPC title