通过建立起资源使用、服务器热点与电源能耗三者的统一优化模型,设计多目标虚拟机动态管理系统,协同优化资源使用、服务器热点和电源能耗,以使得数据中心的效益达到最优.鉴于数据中心负载流量的动态波动性,采用基于滑动窗口的两级状态检测策略来决策迁移虚拟机时刻,基于多目标统一优化模型设计相应算法来解决迁移虚拟机和目的主机的选择问题.最后在Xen虚拟机平台上测试该系统,实验结果表明:本文设计的虚拟机动态管理方法能够较好地协同优化数据中心的资源使用、服务器热点和电源能耗,使得数据中心的效益达到最优.
Relationship among application performance,hot spots and power consumption were studied and dynamic management system for virtual machines was designed by constructing a multi-objective unified optimized model.The focus was on the VM(virtual machine)placement problem which was posed as a multi-objective optimization problem of simultaneously maximizing application performance,minimizing power consumption and eliminating hot spots.A two-level detection strategy based on sliding window was proposed for decide when to migrate a VM.Multi-objective unified optimized models and approaches were presented to handle which and where to migrate a VM.Finally,the system proposed was evaluated by comprehensive experiments based on virtual machine monitor Xen and the results show that it can simultaneously coordinate minimizing total performance loss,thermal dissipation costs and power consumption.