面向业务动态变化的虚拟机迁移技术研究

Comprehensive stable VM migration techniques for managing time-varied workload

针对数据中心内业务需求动态变化下虚拟机的迁移问题,提出了一种基于动态资源需求相关性的虚拟机迁移算法。该算法首先对各个虚拟机的资源需求变化以及服务器资源利用率的动态特性之间的相关性进行定量评估,再根据相关性最优匹配原则以及虚拟机的服务水平协议(SLA)以确定各个待迁移虚拟机的最佳迁移策略。通过与其他虚拟机迁移算法比较,结果表明该算法能更有效地降低动态环境下虚拟机的迁移成本,提高服务器资源利用率以及降低数据中心的能耗。该算法能较好地适用于资源需求动态变化环境下的虚拟机迁移,达到物理资源的高效利用。

This paper developed a novel virtual machine （VM） migration algorithm to solve the VM migration problem in the dynamic environment based on the correlation of variable resource requirements. The algorithm quantified the correlations of changes in the number of resources required by different VMs and the dynamic characteristics of resource usage in servers. Then it determined the optimal migration strategies for VMs to be migrated based on the optimal correlation matching strategy under the consideration of the service level agreement （ SLA ） requirements of VMs. Compared to previous VM migration schemes, the proposed algorithm can not only substantially decrease the migration cost, but also improve the utilization of physi- cal servers and reduce the energy consumption in data centers. It reveals that the algorithm can be applicable for the VM mi- gration as the resource requirements changes dynamically over time, and achieve the efficient usage of physical resources.