云数据中心基于阈值的虚拟机迁移节能调度算法

Threshold-based energy-efficient VM scheduling in cloud datacenters

针对采用虚拟机迁移技术在云计算数据中心同时考虑能耗与服务等级协议（SLA）违约优化的问题,提出一种基于阈值的虚拟机部署节能算法（THRSD-MMT）.该算法通过对虚拟机运行状态的统计分析,估算虚拟机所需中央处理器（CPU）性能需求（每秒处理百万条指令数MIPS）的期望值与标准差,进而动态地计算主机所需MIPS数量;同时,算法结合静态阈值的设置,以便更准确地预测主机的违约情况并判断虚拟机迁移的时机,从而能够在降低SLA违约的同时减少能耗.实验结果表明：与其他算法相比,提出的算法能够显著降低SLA违约率并节能,具有较好的综合性能.

To address the problem that utilizing the VM（virtual machine） live migration technique to simultaneously reduce the energy consumption and the SLA（service level agreements） violation ratio,an energy-efficient VM deployment algorithm that based on a dynamic prediction policy and a static utilization thresholds（THRSD-MMT） was proposed.By the statistical data analyzing of the historical running state of the VMs,the proposed algorithm firstly calculated the expectations and the standard deviations of the central processing unit（CPU） capacity（MIPS） requested by the VMs.Then,the dynamic CPU capacity（MIPS） requested by each host was dynamically evaluated.Furthermore,by integrating dynamic estimation and the static threshold,the algorithm was able to predict the potential SLA violation and determine when the VMs could be migrated more accurately,such that the energy consumption and the SLA violation ratio could both be optimized.The experimental results demonstrate that comparing with other algorithms,and the proposed algorithm can significantly reduce the SLA violation ratio and achieve better comprehensive performance.