• 论文 •    

面向高可信集群计算的能量有效任务分配

江维,熊光泽,常政威,丁旭阳,桑楠   

  1. 电子科技大学 计算机科学与工程学院,四川成都610054
  • 出版日期:2009-12-15 发布日期:2009-12-25

Energy-efficient task allocation for high-dependable cluster computing systems

JIANG Wei, XIONG Guang-ze, CHANG Zheng-wei, DING Xu-yang, SANG Nan   

  1. School of Computer Science & Engineering, University of Electronic Science & Technology of China, Chengdu 610054, China
  • Online:2009-12-15 Published:2009-12-25

摘要: 为解决集群计算系统的失效和高能耗问题,提出了一种灵活的高可信集群计算任务调度框架。该框架基于动态电压调整、任务恢复执行和完美接纳控制机制,设计了一种可靠性感知和能量感知的实时任务全局分配算法。结合局部松驰回收机制,在维持任务可靠性目标的情况下,该算法可进一步降低系统能耗和任务拒绝率。实验表明,该算法在能量节省、可靠性和任务拒绝率方面具有优越性。

关键词: 集群计算, 高可信, 能量有效, 可靠, 实时调度

Abstract: A flexible task scheduling framework was proposed to deal with failure and high energy consumption problems in high-dependable cluster computing systems. Based on dynamic voltage scaling, task execution recovery, and perfect admission control mechanisms, a reliability-aware and energy-aware global allocation algorithm was designed for real-time tasks. In conjunction with local slack reclaim mechanism, the algorithm could further reduce energy consumption and task rejection ratio while maintaining the reliability goal for each task. Simulation results demonstrated superior performances of the proposed algorithm on energy-saving, reliability improvement, and task rejection ratio.

Key words: cluster computing, high-dependable, energy-efficient, reliability, real-time scheduling

中图分类号: