详细信息
负载驱动的机群系统动态功耗管理研究 ( EI收录)
Study of workload-driven dynamic power management for high performance computing clusters
文献类型:期刊文献
中文题名:负载驱动的机群系统动态功耗管理研究
英文题名:Study of workload-driven dynamic power management for high performance computing clusters
作者:梁爱华[1,2];肖利民[1];庞瑜[1];李勇男[1];阮利[1]
第一作者:梁爱华
通讯作者:Xiao, L.
机构:[1]北京航空航天大学计算机学院,北京100191;[2]北京联合大学计算机技术研究所,北京100101
第一机构:北京航空航天大学计算机学院,北京100191
通讯机构:[1]School of Computer Science and Engineering, Beihang University, Beijing 100191, China
年份:2012
卷号:22
期号:11
起止页码:1143-1148
中文期刊名:高技术通讯
外文期刊名:High Technology Letters
收录:CSTPCD;;EI(收录号:20130215883512);Scopus(收录号:2-s2.0-84871914228);北大核心:【北大核心2011】;CSCD:【CSCD2011_2012】;
基金:863计划(2011AA01A205),国家自然科学基金(61003015),教育部博士点课题(20101102110018)和北京市自然科学基金(4122042)资助项目.
语种:中文
中文关键词:动态功耗管理(WDPM);负载驱动;预唤醒;反馈机制;机群
外文关键词:dynamic power management ( WDPM), workload-driven, pre-wakeup, feedback mechanism,cluster
摘要:为了通过动态功耗管理方法解决高性能计算机系统功耗与性能之间的权衡问题,提出了一种适用于同构机群环境下负载驱动的动态功耗管理(WDPM)策略。该WD.PM策略根据负载随时间变化的不均衡性,通过分析并行负载日志的到达特征,在改进超时动态功耗管理的基础上运用了基于负载预测的预唤醒方法和实时反馈修正机制。以真实负载日志和实际系统参数为源数据进行了模拟实验,而且做了WDPM策略与超时策略的对比测试。实验结果表明,WDPM策略能在增加极少量功耗的情况下,有效减少作业平均等待时间和系统中节点状态切换次数,能更好地获得性能与功耗间的平衡。
A workload-driven dynamic power management (WDPM) strategy is proposed to improve the tradeoff between power consumption and performance in homogeneous computing clusters. Through the empirical analysis of real workload logs in production systems, the WDPM strategy integrates the load prediction-based pre-wakeup approach and the dynamic feedback-based revising mechanism based on the improvement of the timeout strategy. By using the data from real workload logs and practical systems, the extensive simulations were conducted to investigate the performance and energy consumption of the strategy. The experimental results indicate that, as compared with the timeout strategy, the WDPM strategy can effectively reduce the average wait time of jobs and the node state switc- hing times of a system with a very little increase of power consumption. Therefore, it can alleviate the performance loss and achieve the better tradeoff of performance and power consumption.
参考文献:
正在载入数据...