摘要
对于虚拟计算系统(如云计算的基础设施即服务)而言,除对外提供简单的单虚拟机外,还需要通过虚拟小组来响应高性能计算任务的服务请求。此时,伴随节点间通信峰值而来的网络拥塞问题是影响计算效率提升的关键点。从可持续发展的角度入手,以高性能计算(HPC)任务在虚拟化平台上的最优部署为目标,通过对称多处理虚拟机的应用,提出了一种基于布局的节能调度方法。仿真实验显示,该方法能够有效缓解虚拟化平台中的通信峰值问题,同时较大幅度地降低高性能计算任务的运行时间。
For virtualized computing system s, e. g. , infrastructure as a service of cloud computing,there is demand from on-demand users to run high performance computing (HPC)jobs network jam always occurs during communication peaks among VM s within a VG , especially whenthey are co-serving a HPC job. Thus, they are challenging problems to alleviate network peaks and to use the network efficiently i terms of power consumption.This paper introduces a layout-based power-efficient co-scheduling algorithm forH PC jobs hosted by virtualized platforms. In these platforms,symmetric multiprocessing virtual machine is recommended against unique processor virtual machine.Emulation indicates that the proposed algorithm out performs others in managing networking peaks ofvirtualized environment and shortening execution tme of HPC jobs.
出处
《上海电机学院学报》
2016年第3期164-169,186,共7页
Journal of Shanghai Dianji University
基金
上海高校青年教师培养资助计划资助(14AZ23)
上海电机学院科研启动金项目资助(13QD01)
上海电机学院重点学科资助(13XKJ01)
关键词
高性能计算任务
云计算
虚拟化
节能调度
high performance computing jo b
cloud computing
virtualization
energy-saving scheduling