Adaptive dynamic buffering system for power management in server clusters

• US 7,441,135 B1
• Filed: 01/14/2008
• Issued: 10/21/2008
• Est. Priority Date: 01/14/2008
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Kephart, Jeffrey O., Levine, David W., Chan, Hoi Y.
• Primary Examiner(s): CONNOLLY, MARK A
• Application Number:
  US12/013,952
• Time in Patent Office: 281 Days
• Field of Search: 713/300, 713/320, 713/323, 713/324
• US Class Current: 713/320
• CPC Class Codes: G06F 1/3203 Power management, i.e. event-based initiation of a power-saving mode
• CPC Class Codes: G06F 1/3246 by software initiated power-off
• CPC Class Codes: G06F 1/3287 by switching off individual functional units in the computer system
• CPC Class Codes: Y02D 10/00 Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

A system, method and computer program product for managing power consumption in a system comprising at least two computing devices. The at least two computing devices are organized as part of active computing device group or a ready computing device group, the ready computing devices configured into various modes of readiness including a standby mode, a hibernating mode, and a power-off mode. A control device receives obtained sensed parameters indicative of the processing load at the group of active computing devices and determines a “Readiness” index value representing a state of system readiness which is a function of a transition time parameter representing a time for moving one or more computing devices from the second group of ready computing devices to the first group of active devices in response to a service level violation. To achieve maximum power saving and yet meet various performance targets, the ready computing devices are dynamically allocated from amongst three states: standby, hibernating or power-off, based on the current readiness index value.

Claims

1. A method for managing power consumption in a system comprising at least two computing devices, the method comprising:

providing one or more computing devices in a first group of active computing devices each adapted for processing a computational workload, and,

sensing a state of processing operations at said active computing devices,

receiving, at a control device, a signal requesting that additional computing devices from a second group of inactive computing devices be added to the first group of active computing devices, or that computing devices from the first group of active computing devices be placed in the second group of inactive computing devices according to said state of processing operations;

determining, by said control device, a system transition time TT value for moving one or more computing devices from the second group of ready computing devices to said first group of active devices, and comparing said TT value against a minimum value, M_min, representing a minimum acceptable delay time for switching a standby computing device from said second group of ready computing devices to said first group of active computing devices and, against a maximum value, M_max, representing a maximum delay interval when all computing devices are in a powered off state;

the control device further dynamically generating a readiness index parameter RI_t at a current time indicating a state of system readiness where RI_t is a real number from 0 to 1, where 0 represents minimum readiness and 1 represents maximum readiness, said RI_t calculated according to one of;

RI_t=RI<sub>t−

1</sub>+(TT−

M_min)/M_max, when increasing system readiness according to a predetermined system response time policy, or,

RI_t=RI<sub>t−

1</sub>−

(TT−

M_min)/M_max, when decreasing system readiness according to a predetermined system response time policy, and, where RI<sub>t−

1</sub> is a RI value prior to said current time; and

,

dynamically allocating, by said control device, one or more computing devices in said second group of ready computing devices in various stages of stand-by, hibernating and power-off readiness according to a value of said readiness index parameter, wherein

if RI=0, all N ready computing devices in the second group are in powered-off state where N=number of total ready computing devices in the second group; and

,

if RI=1, all N ready computing devices in the second group are in one of;

standby state or hibernating state;

wherein, a time it takes to bringing a computing device from said second group of ready computing device to an active readiness state in said first group is minimized, while maximizing energy savings in said system.

2. The method for managing power consumption as claimed in claim 1, wherein said dynamically allocating includes:

providing a number N*(1−

RI) of powered off computing devices in said second group;

providing a number N*RI/2 of standby computing devices in said second group; and

,

providing a number N*RI/2 of hibernating computing devices in said second group.