Research On Virtual Machine Migration Mechanism Aimed At Green Cloud Computing

With the promotion of cloud computing,it is more and more important for environmental protection.More and more attention has being paid to the factor of energy.How to implement the green cloud computing has become one of the research points that need to be studied deeply.Green cloud computing theory is dedicated to reducing the loss of hardware resources,such as servers and refrigeration systems so as to protect the environment.The dynamic migration mechanism of Virtual Machine(VM)in green cloud computing can effectively improve the utilization of cloud data center resources.Appropriate load prediction algorithm and VM scheduling algorithm are important for the realization of VM dynamic migration mechanism.VMs' migration time and their target host can be obtained by accurate load prediction algorithm and reasonable VM scheduling algorithm so as to reduce the the number of VMs' migration and the energy consumption of cloud data center effectively.This dissertation focuses on the research of VM migration mechanism for green cloud computing,especially for load prediction algorithm and VM scheduling algorithm.Following two contents are mainly been studied.(1)A novel VM scheduling algorithm based on the gravitational effect has been proposed to solve the problem that most current VM scheduling algorithms fail to consider the actual physical node deployment scenarios of cloud data centers,which tends to cause local overheating of cloud data centers.Inspired by the physical gravitation model,a gravitation calculation function has been created to design the data center gravitation model.Also,we define the thermal repulsion factor between physical node and VM,and the logical gravitation.What is more,based on the concept of dimension reduction,the algorithm conducts the two-dimensional plane target selection twice to reduce the computational cost.Experiment shows that the algorithm can significantly reduce the energy consumption of cloud computing system,and enhance the stability of the system so as to be more suitable for current cloud data centers.(2)Considering the shortcomings of the existing load forecasting algorithm,a load prediction algorithm has been proposed according to the characteristics of load data which combined the advantages of Cloud Model and Markov chain for optimizing the dynamic migration mechanism.The algorithm put forward the transfer matrix of Markov Chain and weighted summation method to obtain the forecast data instead of traditional calculation of membership degree of Cloud Model,and proposed the way of multi-dimension for the algorithm.The experimental result shows that the proposed algorithm can obtain higher prediction accuracy than the traditional algorithm,which is suitable for the load prediction in cloud computing environment.