Cloud computing attracted great attention in both industry and research communities for the sake of its ubiquitous, elasticity and economic services. The first class concern of cloud providers is power management for both reducing their total cost of ownership and green computing objectives. To reach the goal, a system framework is presented which has different modules. The main concentration of the paper is on virtual machine (VM) consolidation module which launches users requested VMs on the minimum number of active servers to reduce datacenter total power consumption (TPC). In this paper, the VMs consolidation is abstracted to two-dimensional bin-packing problem and also is formulated to an integer linear programming. Since the papers in the literature scarcely are aware of skewness in resources of requested VMs and for discrete nature of search space, this paper presents the resource skewness-aware VMs consolidation algorithm based on improved thermodynamic simulated annealing approach because resource skewness potentially compels the algorithm to activate additional servers. The proposed SA-based algorithm is validated in extensive scenarios with different resource skewness in comparison with two heuristics and two meta-heuristics. The average results reported from different scenarios proves superiority of proposed algorithm in comparison with other approaches in terms of the number of used servers, TPC, and total resource wastage of datacenter.

云计算因其无处不在，弹性和经济的服务而吸引了业界和研究界的极大关注。云提供商的首要关注点是电源管理，它既可以降低总体拥有成本，又可以实现绿色计算目标。为了实现该目标，提出了具有不同模块的系统框架。本文的主要内容是在虚拟机（VM）整合模块上，该模块可在最少数量的活动服务器上启动用户请求的VM，以减少数据中心的总功耗（TPC））。在本文中，虚拟机合并被抽象为二维装箱问题，并且被表述为整数线性规划。由于文献中的论文几乎不了解请求的VM的资源偏度以及搜索空间的离散性，因此本文提出了一种基于改进的热力学模拟退火方法的可感知资源偏度的VM整合算法，因为资源偏度可能迫使该算法激活其他服务器。与两种启发式算法和两种元启发式算法相比，所提出的基于SA的算法在具有不同资源偏斜的广泛场景中得到了验证。从不同情况下报告的平均结果证明，与使用其他方法的服务器数量相比，该算法具有优越性。TPC和数据中心的总资源浪费。