Research On Energy Saving Based Virtual Machine Deployment And Consolidation

In recent years,with the widespread application of cloud computing technology,the energy consumption of the cloud data center has been receiveing more and more attention.Meanwhile,with the rise of green computing concept,the virtual machine deployment and consolidation algorithm which is energy efficient has an important significance.The thesis is based on those two aspects: the deployment of virtual machine in data center and the consolidation of virtual machine during the operationing of data center.To solve those problems,this thesis proposes two algorithms: a virtual machine deployment algorithm,which considers the energy consumption of network equipments;a virtual machine consolidation algorithm based on workload prediction.The main work are as follows:(1)For the deployment of virtual machines,traditional algorithms do not consider the energy consumption of network equipments.This thesis proposes a Modified Particle Swarm Optimization algorithm(MPSO),MPSO is different from traditional PSO: MPSO adopts a two-dimensional encoding scheme,the position of a particle represents a solution of the virtual machine deployment problem;more over,MPSO redefining a new fitness function,which considers the total energy consumption;finally,MPSO adopts an local fitness strategy to update the local position of the particle,which is energy-aware and different from the random selection strategy of the traditional PSO.The experiments show that MPSO can reduce the energy consumption of data center when compared with traditional methods.(2)For the consolidation of virtual machine.This thesis proposes a Virtual Machine Consolidation algorithm based on Load Prediction(VMCLP).The traditional virtual machine consolidation algorithm determine the physical machine overload or lowload by the current state of the physical machine,however,this methods can't reflect the state of the physical machine in the future.VMCLP adopts markov chain to predict the load of each physical machine,the physical machine's mode is determined by the the current load and the predict load,according to VMCLP,wether to migrate virtual machines is depend on both of them.Finally,VMCLP adopts a minimum migration cost strategy to improve the virtual machine selection strategy and a best fit virtual machine strategy to improve the virtual machine placement strategy.Experimental results indicate that VMCLP can reduce not only the energy consumption,but also the migrations of virtual machine.