作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2010, Vol. 36 ›› Issue (24): 96-98. doi: 10.3969/j.issn.1000-3428.2010.24.034

• 网络与通信 • 上一篇    下一篇

基于在线负载预测的动态集群节能配置策略

刘 斌1,杨 坚1,赵 宇2   

  1. (1. 中国科学技术大学自动化系,合肥 230027;2. 中国科学院上海微系统与信息技术研究所,上海 200050)
  • 出版日期:2010-12-20 发布日期:2010-12-14
  • 作者简介:刘 斌(1984-),男,硕士研究生,主研方向:服务器集群节能,无线通信;杨 坚,副教授、博士;赵 宇,副研究员
  • 基金资助:

    国家“863”计划基金资助项目(2006AA01Z114);国家自然科学基金资助项目(60802037);新世纪优秀人才支持计划基金资助项目(NCET-08-0522);中科院优秀博士论文获得者启动基金资助项目

Dynamic Cluster Configuration Strategy for Energy Conservation Based on Online Load Prediction

LIU Bin1, YANG Jian1, ZHAO Yu2   

  1. (1. Department of Automation, University of Science and Technology of China, Hefei 230027, China; 2. Shanghai Institute of Micro-system and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China)
  • Online:2010-12-20 Published:2010-12-14

摘要:

动态集群配置中的调节方式大多基于特定的物理实验模型而非数学模型描述。针对上述问题,提出基于预测的动态集群配置策略,根据网络中服务请求的历史信息,采用最小均方误差预测未来时刻服务请求情况,根据负载请求与集群处理能力决定服务器规模的增减,动态调节服务器集群中计算机的开启与关断。实验结果证明了该调度策略的可行性和优越性。

关键词: 服务器集群, 动态集群配置, 预测算法, LMS算法, 节能

Abstract:

Previous dynamic cluster configuration methods are based on the specific physical experimental models without the description of mathematical models. Aiming at the problem, this paper proposes a prediction-based dynamic clusters configuration strategy, which uses least mean square to predict the situation of service requests in the future time according to the network historical information of service requests. On the basis of the load requests and the clusters processing power, it decides the servers’ scale and dynamically adjusts the opening and shutdown of the computers in the server cluster. Experimental result verifies the feasibility and superiority of the schedule strategy.

Key words: server cluster, dynamic cluster configuration, prediction algorithm, LMS algorithm, energy conservation

中图分类号: