关键词: 云计算, 虚拟机整合, 高斯混合模型, 主机过载概率, 服务质量

Abstract: High efficient virtual machine consolidation can deploy virtual machines reasonably and reduce energy consumption while guaranteeing quality of service. However, due to the uncertainty of resource requirements by virtual machines, traditional consolidation methods cannot make the deployment be stable for a long time, that influence the efficiency of consolidation. Considering this problem, gaussian mixture model was used to establish a stochastic load model and the model parameters estimated by the exception maximization algorithm. The overload probability was estimated by its load model for each host, and by which determine whether the host is overloaded or not. If one host is overloaded, virtual machines with less memory that can significantly degrade overload probability were migrated preferentially. The migrated virtual machine was replaced in the host whose overload probability is minimal after accepting the migrated virtual machine. The experimental results indicate that the VMC method based on GMM can reduce energy consumption in data center and improve quality of service.

Key words: cloud computing, virtual machine consolidation, gaussian mixture model, overload probability of hosts, quality of service