基于GSPN的数据中心动态能耗管理研究

发布时间：2018-05-01 13:38

本文选题：数据中心 + 动态能耗管理　；参考：《系统仿真学报》2014年12期

【摘要】：随着云数据中心的大规模应用,其高能耗、高污染等问题日益突出。围绕数据中心的动态能耗管理(DEM,Dynamic Energy Management)问题,提出了一种基于广义随机Petri网(GSPN,Generalized Stochastic Petri Nets)的DEM建模、求解与分析的系统性能评价框架。该框架综合考虑了多种任务调度策略的联合应用与服务器工作模式的动态调整,设计了合理的能耗优化策略,在保证系统服务质量(Qo S,Quality of Service)的同时,降低了数据中心能耗。仿真实验表明,所建立的性能评价框架,在不同系统负载下,可实现最佳任务调度策略的动态选择和服务器工作模式的动态切换,相较于随机调度策略,所提出的均衡调度策略和集中调度策略,分别可降低28%和43%的能耗。
[Abstract]:With the large-scale application of cloud data centers, the problems of high energy consumption and high pollution have become increasingly prominent. In this paper, a system performance evaluation framework based on generalized stochastic Petri network (GSPN / generalized Stochastic Petri Nets) is proposed to model, solve and analyze the dynamic Energy management of data center. The framework synthetically considers the joint application of multiple task scheduling strategies and the dynamic adjustment of the server working mode, and designs a reasonable energy consumption optimization strategy, which can ensure the quality of service of the system and reduce the energy consumption of the data center at the same time. Simulation results show that the proposed performance evaluation framework can realize dynamic selection of optimal task scheduling strategy and dynamic switching of server working mode under different system loads, compared with random scheduling strategy. The proposed balanced scheduling policy and centralized scheduling policy can reduce the energy consumption by 28% and 43% respectively.
【作者单位】：北京信息科技大学计算机学院网络文化与数字传播北京市重点实验室;
【基金】：国家自然科学基金(61370065)
【分类号】：TP308;TP301.1

【相似文献】