Server configured for managing power and performance

• US 20080104430A1
• Filed: 10/31/2006
• Published: 05/01/2008
• Est. Priority Date: 10/31/2006
• Status: Active Grant

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Vaden, Thomas L., Malone, Christopher G., Goldstein, Martin, Huscroft, Carey
• Granted Patent: US 8,001,407 B2
• Time in Patent Office: Days
• 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 server manages power and performance. The server comprises a plurality of processor and associated storage and input/output interface components, arranged into a plurality of independent compute cells. One or more performance metrics descriptive of performance of the independent compute cell plurality are selected and an optimization evaluation of the one or more performance metrics is performed. The independent compute cell plurality is tuned based on the optimization evaluation.

Claims

1. A method of managing power and performance in a server comprising:

operating a server comprising a plurality of processor and associated storage and input/output interface components, arranged into a plurality of independent compute cells;

selecting one or more performance metrics descriptive of performance of the independent compute cell plurality;

performing an optimization evaluation of the one or more performance metrics; and

tuning the independent compute cell plurality based on the optimization evaluation.

2. The method according to claim 1 further comprising:

selecting one or more performance metrics from a group consisting of performance/Watt, maximized system performance, server energy minimization.

3. The method according to claim 1 further comprising:

tuning the independent compute cell plurality based on application type of applications executing on the independent compute cell plurality.

4. The method according to claim 1 further comprising:

tuning the independent compute cell plurality selected from a group consisting of;

enabling use of a control that causes a processor to halt execution whereby one or more processors in the independent compute cell plurality enter a low power state when idle;

clock throttling for a processor clock in one or more processors in the independent compute cell plurality;

dynamically transitioning on-demand processor performance state (P-state); and

adjusting processor multiplier and/or requested voltage level in real-time.

5. The method according to claim 1 further comprising:

priority-ranking applications executing on the independent compute cell plurality; and

stopping and/or reducing performance of components in the independent compute cell plurality that execute applications ranked with low priority.

6. The method according to claim 5 further comprising:

triggering the priority-ranking according to a condition selected from a group of conditions consisting of;

a predefined abnormal ambient temperature condition;

a predefined data center abnormal power condition; and

a predefined result of an applied business rule that minimizes Information Technology (IT) cost.

7. The method according to claim 5 further comprising:

applying priority-ranking of applications and stopping and/or reducing performance of components across multiple compute cells and/or systems in a data center.

8. The method according to claim 1 further comprising:

detecting a condition that activates an application-throttling function; and

activating the application-throttling function that operates in a low-power/moderate-response state comprising;

evoking a spin-lock state for one or more selected processors in the independent compute cell plurality.

9. The method according to claim 8 further comprising:

assigning the application-throttling function an execution priority that is higher than currently executing applications.

10. The method according to claim 8 further comprising:

evoking the low-power/moderate-response state in response to a condition selected from a group consisting of;

a direct user input signal;

a detected power condition and/or a detected temperature condition; and

an external stimulus.

11. The method according to claim 8 further comprising:

selecting a percentage of processor time for which the one or more selected processors is to operate in the spin-lock state; and

operating the one or more selected processors in the spin-lock state for the selected percentage of processor time.

12. The method according to claim 1 further comprising:

operating the server and associated independent compute cell plurality as a plurality of subsystems;

determining a target reduction in power consumption and/or temperature in the server; and

terminating operations of selected subsystems in the independent compute cell plurality to maintain the target reduction.

13. The method according to claim 12 further comprising:

identifying redundant subsystems of the independent compute cell plurality; and

terminating operations of the redundant subsystems in the independent compute cell plurality to maintain the target reduction.

14. The method according to claim 12 further comprising:

operating the server and the independent compute cell plurality as a plurality of subsystems selected from a group consisting of interface subsystems, input/output subsystems, cooling subsystems, control subsystems, and manageability subsystems.

15. The method according to claim 12 further comprising:

determining a thermal envelope in the server; and

terminating operations of selected subsystems in the independent compute cell plurality to maintain the thermal envelope.

16. The method according to claim 12 further comprising:

determining a thermal envelope in the server; and

varying fan speeds to maintain the thermal envelope.

17. A server comprising:

a plurality of processor and associated storage and input/output interface components arranged into a plurality of independent compute cells;

a controller coupled to the independent compute cell plurality configured select one or more performance metrics descriptive of performance of the independent compute cell plurality, perform an optimization evaluation of the one or more performance metrics, and tune the independent compute cell plurality based on the optimization evaluation.

18. The server according to claim 17 further comprising:

the controller configured for;

selecting one or more performance metrics from a group consisting of performance/Watt, maximized system performance, server energy minimization;

tuning the independent compute cell plurality selected from a group consisting of;

enabling use of a control that causes a processor to halt execution whereby one or more processors in the independent compute cell plurality enter a low power state when idle;

clock throttling for a processor clock in one or more processors in the independent compute cell plurality;

dynamically transitioning on-demand processor performance state (P-state); and

adjusting processor multiplier and/or requested voltage level in real-time.

19. The server according to claim 17 further comprising:

the controller configured for;

priority-ranking applications executing on the independent compute cell plurality; and

stopping and/or reducing performance of components in the independent compute cell plurality that execute applications ranked with low priority;

triggering the priority-ranking according to a condition selected from a group of conditions consisting of;

a predefined abnormal ambient temperature condition;

a predefined data center abnormal power condition; and

a predefined result of an applied business rule that minimizes Information Technology (IT) cost.

20. The server according to claim 17 further comprising:

the controller configured for;

detecting a condition that activates an application-throttling function;

assigning the application-throttling function an execution priority that is higher than currently executing applications; and

activating the application-throttling function that operates in a low-power/moderate-response state comprising;

evoking a spin-lock state for one or more selected processors in the independent compute cell plurality;

evoking the low-power/moderate-response state in response to a condition selected from a group consisting of;

a direct user input signal; and

an external stimulus;

selecting a percentage of processor time for which the one or more selected processors is to operate in the spin-lock state; and

operating the one or more selected processors in the spin-lock state for the selected percentage of processor time.

21. The server according to claim 17 further comprising:

the controller configured for;

operating the server and the independent compute cell plurality as a plurality of subsystems selected from a group consisting of interface subsystems, input/output subsystems, cooling subsystems, control subsystems, and manageability subsystems;

determining a target reduction in power consumption and/or temperature in the server; and

terminating operations of selected subsystems in the independent compute cell plurality to maintain the target reduction.

22. A control apparatus adapted for managing power and performance in a server comprising:

a server comprising a plurality of processor and associated storage and input/output interface components, arranged into a plurality of independent compute cells;

means for selecting one or more performance metrics descriptive of performance of the independent compute cell plurality;

means for performing an optimization evaluation of the one or more performance metrics; and

means for tuning the independent compute cell plurality based on the optimization evaluation.