绿色数据中心流量感知虚拟机迁移(英文)

Towards Energy Saving in Data Centers Using Traffic-aware VM Migration

绿色数据中心内将虚拟机迁移并整合到更少数目的物理机上,从而关闭或睡眠空闲的物理机,是非常高效的数据中心节能方法.本文构建了新的流量感知虚拟机迁移模型,把物理机使用数量、流量代价、迁移代价影响都置入模型,使得虚拟机的迁移方案达到缩减物理机目的的同时,减少网络代价和迁移过程中的代价。提出了最小流量代价优先匹配的贪婪算法来解决这个多目标优化问题。仿真结果显示,这个多目标的优化中,多个优化目标之间存在一定的冲突,因此需要折中处理。该算法在物理机数量和迁移代价两个目标上做出了微小的让步,得到了较好的流量代价优化。实际的迁移算法应用中,还应考虑各个部分的真实能耗,采取合理的折中处理.

Powering off or sleeping the idle servers based on workload consolidation and virtual machine migration is very efficienct on energy saving in green data center.In this article,we take three objectives into consideration and propose a completely new model of traffic-aware virtual machine migration.The model lead to find new embedding blueprint for virtual machines,which would minimize servers consumption,network traffic cost and the migration cost simultaneously.Then,we introduce a traffic-aware greedy algorithm called minimum traffic cost first fit to deal with the multi-objective optimization problem.Simulation results shows that there should be a tradeoff among the three objectives and our greedy algorithm compromise a little on the servers consumption and migration cost making for a better traffic cost.However,the compromise needs to be reconsidered in practice,depending on the actual energy consumption of each part.