Power management for clustered computing platforms

• US 7,051,215 B2
• Filed: 06/13/2003
• Issued: 05/23/2006
• Est. Priority Date: 06/13/2003
• Status: Active Grant

• Current Assignee: Sony Corporation Of America (Sony Group Corp.)
• Original Assignee: Intel Corporation
• Inventors: Zimmer, Vincent J., Rothman, Michael A.
• Primary Examiner(s): ELAMIN, ABDELMONIEM I
• Application Number:
  US10/461,038
• Publication Number:
  US 20040255171A1
• Time in Patent Office: 1,075 Days
• Field of Search: 713/300, 713/320, 713/323, 713/324
• US Class Current: 713/300
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode

Abstract

Methods, apparatus, and systems for power management of clustered computing platforms. Firmware provided on each platform is loaded for operating system runtime availability. The firmware traps power change events corresponding to a change in a power state of a component or device associated with a platform, such as a processor or a peripheral device. In response to the power change event, the firmware sends corresponding data to a power arbiter using an out-of-band (OOB) communication network/channel. The power arbiter determines an appropriate reallocation of power consumption allotments for the various platforms under its control based on a power management policy. Updated power consumption allocation information is then sent to selected platforms via the OOB network/channel. Upon receipt of the updated allotments, corresponding registers and/or tables are updated on the platforms to effectuate a change in power consumption of one or more components/devices associated with the platform.

Claims

1. A method for managing power consumption across a plurality of computing platforms, comprising:

monitoring power change events corresponding in a change in a power state of a component or device associated with a given platform;

reallocating a power consumption allotment for at least one computing platform in response to the power change event in accordance with a power management policy; and

effectuating the reallocated power consumption allotment on said at least one computing platform in a manner that is transparent to an operating system running on said at least one computing platform.

2. The method of claim 1, wherein the plurality of computing platforms comprise a group of server blades operating in a blade server environment.

3. The method of claim 2, wherein the power change event corresponds to one of insertion of a new blade into a blade server chassis or removal of an existing blade from a blade server chassis.

4. The method of claim 2, wherein the power management policy is implemented across one of a group of server blades, an individual blade server chassis, a plurality of blade server chassis, a blade server rack, or a plurality of blade server racks.

5. The method of claim 1, wherein the method is performed in a manner that is transparent to operating systems running on the computing platforms.

6. The method of claim 1, wherein the method is facilitated by firmware running on each of the plurality of computing platforms, each computing platform including a processor on which an operating system is run, the firmware for that computing platform being executed in a hidden execution mode of the processor.

7. The method of claim 1, wherein the operations of monitoring power change events and reallocating power consumption allotments include:

detecting a power change event corresponding to a first computing platform;

determining power state configuration information pertaining to the power change event

passing the power state configuration information to a power management entity that is external to the first computing platform; and

determining the reallocation of the power consumption allotments at the power management entity.

8. The method of claim 7, wherein the power management entity comprises one of a dedicated management component or remote server coupled in communication to each of the plurality of computing platforms.

9. The method of claim 7, wherein the power management entity comprises a second computing platform.

10. The method of claim 1, wherein the operation of effectuating the reallocated power consumption allotment comprises:

sending updated power consumption indicia to said at least one computing platform;

updating at least one of a register and configuration table on each of said at least one computing platform in response to the updated power consumption indicia to cause a change in a power state of at least one component or device on that computing platform.

11. The method of claim 10, wherein the power consumption indicia comprises a performance state (P-state) identifying a corresponding power consumption state of said at least one component or device.

12. The method of claim 1, wherein the power change event results from a change in a power state of a component or device associated with a computing platform that is invoked by an operating system running on that computing platform.

13. The method of claim 1, wherein power change event results from a change in a power state of a component or device associated with a computing platform that is invoked internally by computing platform hardware.

14. The method of claim 1, wherein the power change event results from a change in a power state of a peripheral device.

15. The method of claim 1, wherein the power management policy is implemented as a peer-based policy under which power consumption allotments are negotiated among multiple computing platforms based on input from those computing platforms.

16. The method of claim 1, wherein the method is facilitated by passing data among the plurality of computing platforms using an out-of-band communication network or channel that operates transparently from operating systems running on the computing platforms.

17. A method for managing power consumption across a plurality of server blades, comprising:

monitoring power change events corresponding in a change in a power state of a component or device associated with a first server blade;

sending power state data indicative of the power change event via a first out-of-band (OOB) communication from the server blade to a power management entity;

reallocating power consumption allotments for selected server blades from among the plurality of server blades in response to the power change event in accordance with a power management policy; and

sending updated power consumption allotments to the selected server blades via a second OOB communication,

wherein the first and second OOB communications are effectuated via firmware executed on a respective processor for each of the plurality of server blades, the processor for a given server blade hosting an operating system to which the OOB communications are transparent.

18. The method of claim 17, wherein the power management policy is implemented as a peer-based policy under which power consumption allotments are negotiated among the plurality of server blades based on input from the plurality of server blades.

19. An article of manufacture comprising a machine-readable medium having instructions stored thereon, which when executed on a computing platform causes the computing platform to perform operations including:

monitor power change events corresponding in a change in a power state of a component or device associated with the computing platform;

send power state data indicative of the power change event to a power management entity;

receive updated power consumption allotment data from the power management entity; and

effectuate a change in a power consumption allotment of at least one component or device associated with the computing platform in response to the updated power consumption allotment data,

wherein the operations of monitoring power change events, sending power state data, and receiving undated power consumption allotment data are performed in a manner that is transparent to an operating system running on the computing platform.

20. The article of manufacture of claim 19, wherein the instructions comprise firmware instructions.

21. The article of manufacture of claim 19, wherein the computing platform comprises a server blade.

22. A blade server system, comprising:

a chassis, including a plurality of slots in which respective server blades may be inserted;

an interface plane having a plurality of connectors for mating with respective connectors on inserted server blades and providing communication paths between the plurality of connectors to facilitate in out of band (OOB) communication channel; and

a plurality of server blades, each including a processor and firmware executable thereon to cause that server blade to perform operations including;

monitor power change events corresponding in a change in a power state of a component or device associated with the server blade;

send power state data indicative of the power change event to a power management entity via the OOB communication channel;

receive updated power consumption allotment indicia from the power management entity via the OOB channel; and

effectuate a change in a power consumption allotment of at least one component or device associated with the server blade in response to the updated power consumption allotment indicia,

wherein the operations of monitoring power chance events, sending power state data, receiving undated power consumption allotment data, and effectuating a change in the power consumption allotment are performed in a manner that is transparent to an operating system running on the computing platform.

23. The blade server system of claim 22, further comprising a management card coupled to the interface plane, said management card comprising the power management entity.

24. A blade server system of claim 22, wherein the power management entity is embodied on at least one of the plurality of server blades.

25. The blade server system of claim 24, wherein the power management entity enforces a power management policy via execution of firmware on said at least one of the plurality of server blades, said power management policy defined to be implemented across one of a group of server blades, an individual blade server chassis, a plurality of blade server chassis, a blade server rack, or a plurality of blade server racks.

26. The blade server system of claim 24, wherein the power management policy is implemented as a peer-based policy under which power consumption allotments are negotiated among multiple server blades based on input from the multiple server blades.