Research On Virtual Machines Consolidation Based On Energy And Qos Optimization

Cloud service providers improve the utilization of physical resources and reducing energy consumption by virtualization technology and virtual resource management.However,due to the stochastic requests of cloud users that make the resource allocation cannot always meet the requests easily.This situation will further affect the quality of services.The virtual machines consolidation can improve the resource utilization and reduce energy consumption in data center.This research will study the virtual machines consolidation method in the prediction of workloads,resource competition relationships,and the optimization model of virtual machines placement.The main innovations in this article are as follows:(1)Concerning the problem that the workload of hosts in data center cannot estimate accurately,a Gaussian Mixture Model-based workload distribution model is proposed and a virtual machines consolidation method is proposed based this model.Firstly,Gaussian Mixture Model(GMM)is used to fit the workload history of hosts.Then,the overload probability of hosts is calculated according to the GMM of workload and resource capacity of hosts.Next,the aforementioned overload probability is taken as the criteria to determine whether the host is overloading or not.Besides,study the criterion of migrated virtual machines hosted by overloading hosts.At last,these migrated virtual machines are placed in new hosts which have less effect on overloading risk of workload after placement estimated by GMM.The result of experiments show that the proposed GMM-VMC method can effectively reduce energy consumption,balance the workload of hosts and improve quality of service.(2)By analyzing the main competition in virtual machines placement,modeling the competition relationship between the provider and users in the placement by strategic game theory.Take the cloud service provider(Provider)and an agent(Broker)representing all virtual machines as players.Then,define the set of participants' actions,based on the difference in resources provided by the Provider and the difference in resources that the Broker applies for the virtual machines.Finally,calculating payment values of Provider and Broker by the virtual machines placement map of each strategy groups in the game matrix,then get the nash equilibrium solution of the game.Experiments show that the proposed SG-VMC method can get the balance between reducing energy consumption and improve quality of service.(3)A virtual machine placement optimization model is established for virtual machine placement problems,and a discrete differential evolution algorithm model is proposed to solve the model.Firstly,improve the differential evolution algorithm,and then proposes a virtual machine placement strategy and method based on the improved discrete differential evolution algorithm.In the further research,combinedwith host overloading detection and virtual machine selection algorithms,an energy and QoS-aware virtual machines consolidation method is proposed.Experiments show that compared with other meta-inspired virtual machine consolidation methods,the EQ-VMC method proposed in this paper has considerable advantages in reducing energy consumption in the data center,and shows excellent performance in ensuring the quality of service.