Abstract: Aiming at dynamical changes of application workload requirements, a self-adaptive virtual machine optimal placement strategy is presented. First, through the hotspot detection based on robust local weight regression, this algorithm can decide the overload time of hosts dynamically according to the historical resource occupation information of application workload. Second, through the optimal migration period MPM and minimal migration number MNM, it can choose migrated VMs on overloaded hosts. Finally, the bin-packing algorithm PBFDH based on power consumption aware is used to deploy VMs again. Experimental results show that it not only can reduce energy consumption, but decrease SLA violation.

Key words: cloud computing, energy-efficiency, Service Level Agreement（SLA）, virtual machine placement, robust local weighted regression

摘要： 针对云计算应用负载需求的动态变化特性，提出了一种自适应虚拟机优化部署策略。算法通过基于强局部加权回归的热点发现机制，可以根据负载所体现的资源占用历史信息动态决策主机的超载时机;通过迁移周期最优算法MPM和迁移量最少算法MNM进行超载主机的迁移虚拟机选择;提出基于功耗感知的PBFDH算法对迁移虚拟机再次优化部署。实验结果表明，算法不仅可以降低能耗，还可以降低SLA违例率。

关键词: 云计算, 能效, 服务等级协议, 虚拟机部署, 强局部加权回归