基于成本感知的虚拟机迁移决策算法

Virtual Machine Migration Decision Algorithm on Basis of Cost-sensitive Modeling

在"混合型负载"数据中心的虚拟机迁移过程中,针对主机性能和能源额外损耗问题,提出了一种基于迁移成本感知的迁移决策算法。为减少频繁迁移对数据中心的扰动,设计了利用自回归模型预测物理主机资源利用率并结合双阈值的迁移触发策略,建立了迁移性能和能耗成本评估模型,确定了待迁移物理主机上虚拟机的迁移优先级。仿真实验结果表明,该算法能够在保证服务级别协议不被破坏的条件下,降低"混合型负载"数据中心的迁移次数和能耗。

In view of the host performance and the extra energy consumption during the process of virtual machine migration in the“hybrid load”data center,a migration decision algorithm based on migration cost-sensitive modeling was proposed.In order to reduce the perturbation of frequent migration to the data center,a migration trigger strategy was put forward,which utilizes the autoregressive model to predict the resource utilization of the physical host and combines the two thresholds method.Then,the migration performance and energy cost evaluation model was proposed,which defines the migration priority of virtual machines on the physical hosts to be migrated.The simulation results show that the proposed algorithm can reduce the number of migrations and energy consumption in the data center with the service level agreement guaranteed.