Abstract: The dynamic migration and integration technology of virtual machine is an effective method to reduce the energy consumption of large scale heterogeneous cloud data centers. This paper adopts the means of the load detection based on exponential smoothing method, and then selects the virtual machines to be reallocated based on the principle of minimum transfer time （MMT）, and designs a Power Aware Best Fit Decreasing-Simulated Annealing algorithm （PABFD-SA） to reallocate the virtual machines. This algorithm adopts the physical host sequence obtained by BFD algorithm to be the initial solution to the SA algorithm, and introduces the function of reservation and updates the optimal solution in the search process. The simulation result demonstrates that the algorithm can reduce the total energy consumption of heterogeneous cloud computing system and SLA violations.

Key words: cloud computing, virtual machine consolidation, dynamic migration, power aware, simulated annealing

摘要： 虚拟机动态迁移整合技术是大规模异构云数据中心降低能耗的有效方法。采用指数平滑预测法进行负载检测，然后以最小迁移时间算法（MMT）为原则筛选出待重分配的虚拟机，并就重分配过程中的能耗优化问题设计了一种感知能耗的最佳适配递减和模拟退火组合算法PABFD-SA（Power Aware Best Fit Decreasing-Simulated Annealing）。该算法将BFD算法获取的物理主机序列作为SA算法的初始解，并在搜索过程中加入了保留和更新历史最优解的功能。仿真结果表明，该算法在减少异构云计算系统的总能耗，降低SLA违约方面有一定改善。

关键词: 云计算, 虚拟机整合, 动态迁移, 感知能耗, 模拟退火