华南理工大学学报(自然科学版) ›› 2007, Vol. 35 ›› Issue (9): 60-64.

• 计算机科学与技术 • 上一篇    下一篇

动态电源管理随机模型算法的设计与实现

刘发贵 麦伟鹏 黄凯耀   

  1. 华南理工大学 计算机科学与工程学院,广东 广州 510640
  • 收稿日期:2006-08-01 出版日期:2007-09-25 发布日期:2007-09-25
  • 通信作者: 刘发贵(1963-) ,女,教授,主要从事操作系统与嵌入式软件方面的研究. E-mail:fgliu@ scut. edu. cn
  • 作者简介:刘发贵(1963-) ,女,教授,主要从事操作系统与嵌入式软件方面的研究.
  • 基金资助:

    国家"863" 计划重大软件专项(2004AA1Z2400) ;粤港关键领域重点突破项目(2005A10207005 ,信产厅2004-0005)

Design and Implementation of Stochastic Model Algorithm for Dynamic Power Management

Liu Fa-gui  Mai Wei-peng  Huang Kai-yao   

  1. School of Computer Science and Engineering , South China Univ. of Tech. , Guangzhou 510640 , Guangdong , China
  • Received:2006-08-01 Online:2007-09-25 Published:2007-09-25
  • Contact: 刘发贵(1963-) ,女,教授,主要从事操作系统与嵌入式软件方面的研究. E-mail:fgliu@ scut. edu. cn
  • About author:刘发贵(1963-) ,女,教授,主要从事操作系统与嵌入式软件方面的研究.
  • Supported by:

    国家"863" 计划重大软件专项(2004AA1Z2400) ;粤港关键领域重点突破项目(2005A10207005 ,信产厅2004-0005)

摘要: 基于随机控制的策略优化算法能有效地解决动态电源管理(DPM) 中电源状态切换的能耗问题,从而获得更优的策略.文中通过为系统建立基于马尔可夫决策过程的随机模型,在DPM 框架中实现了DPM 随机模型算法,并对算法进行了实验.结果表明,在不同的性能损耗条件下,可以得到不同的、满足性能要求的优化策略,也就是说,算法在性能和能量损耗间取得了平衡,这也证明了文中介绍的算法实现过程的可行性.

关键词: 随机模型, 马尔可夫链, 动态电源管理, 算法

Abstract:

The policy-optimizing algorithms based on stochastic model can effectively reduce the power consumption of power state transitions for dynamic power management (DPM) and work out a better strategy. In this paper , a stochastic model based on Markov decision processes was established for the DPM system , and the corresponding algorithm was implemented in a material DPM architecture. Then , experiments for the algorithm were carried out. The results indicate that , with the proposed algorithm , different optimized policies satisfying performance requirements can be worked out in different power consumptions , that is , the algorithm strikes a balance between the performance and energy consumption. All this means that the algorithm implementation is feasible.

Key words: stochastic model, Markov chain, dynamic power management, algorithm