Dynamic power management of a computer with self power-managed components

Maryam Triki, Yanzhi Wang, Ahmed C. Ammari, Massoud Pedram

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations

Abstract

This paper presents a Dynamic Power Management (DPM) framework based on reinforcement learning (RL) technique which aims to save power in an Energy-Managed Computer (EMC) system with self power-managed components. The proposed online adaptive DPM technique consists of two layers: component-level and system-level global power manager (GPM). The component-level PM policy is pre-specified and fixed whereas the system-level global PM employs temporal difference learning on Semi-Markov Decision Process (SMDP) for model-free RL, and it is specifically optimized for a multitype application framework. Experiments show that that the proposed HPM scheme enhances power savings considerably while maintaining a good performance level. In comparison with other reference systems, the proposed RL DPM approach performs well under various workloads, can simultaneously consider power and performance and achieves a wide and deep powerperformance tradeoff curves.

Original languageEnglish (US)
Title of host publicationIntegrated Circuit and System Design
Subtitle of host publicationPower and Timing Modeling, Optimization and Simulation - 22nd International Workshop, PATMOS 2012, Revised Selected Papers
Pages215-224
Number of pages10
DOIs
StatePublished - Dec 1 2013
Event22nd International Workshop on Power and Timing Modeling, Optimization and Simulation, PATMOS 2012 - Newcastle upon Tyne, United Kingdom
Duration: Sep 4 2012Sep 6 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7606 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other22nd International Workshop on Power and Timing Modeling, Optimization and Simulation, PATMOS 2012
CountryUnited Kingdom
CityNewcastle upon Tyne
Period9/4/129/6/12

Keywords

  • Dynamic Power Management (DPM)
  • Power optimization
  • Reinforcement Learning (RL)

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Dynamic power management of a computer with self power-managed components'. Together they form a unique fingerprint.

Cite this