Research On Dynamic Integration Strategy Of Virtual Machine Based On MOPSO Algorithm

With the development of cloud computing technology,cloud data center can integrate a large number of computing resources through virtualization technology,and provide stable and high-quality services to Internet users through the network.However,with the growth of market demand and the expansion of the scale of data center,the situation of low utilization of data center resources and load imbalance is becoming more and more serious.How to improve the utilization of data center resources and achieve efficient management of resources has become a great challenge.The main method of cloud data center resource management is the dynamic integration of virtual machines,and the existing virtual machine integration research is mainly carried out from the perspective of reducing the energy consumption of the data center and improving the utilization of certain resources,Ignoring the resource diversity of physical nodes and the load balancing of the data center after migration.In order to improve the resource usage of the cloud data center,achieve the goals of balancing load and improving resource utilization,it is necessary to study the problems in the process of dynamic integration of virtual machines and give corresponding strategies.This thesis analyzes and studies the migration trigger,the selection of the virtual machine to be migrated,and the placement of the virtual machine in the dynamic integration of virtual machines.The main contributions are as follows:1.In response to the issue of of migration trigger mechanism,this thesis proposes a virtual machine migration trigger strategy based on load prediction.First of all,in order to ensure the running performance of the virtual machine,the stability of the virtual machine is guaranteed by setting the resource utilization threshold of the physical node to reserve resources.Then,the adaptive exponential smoothing prediction method is used to predict the load trend of the physical node,and the necessity of the node to adjust the virtual machine is judged based on the prediction result,thereby reducing the number of virtual machine migrations caused by the instantaneous load fluctuation of the data center,and improving the accuracy of virtual machine migration judgment.2.In response to the problem of selecting virtual machines for migration,this thesis proposes a migration virtual machine selection algorithm based on node resource load.First of all,for the problem of selecting overloaded nodes for migration,comprehensively consider the resource diversity of overloaded physical machines,and measure the degree of overload of various resources of physical machines by defining node saturation.Then,based on the virtual machine resource occupancy and the virtual machine migration cost,a method for determining the priority of virtual machine migration is proposed.Finally,the virtual machine migration priority is used to select the virtual machine that occupies a lot of overloaded resources and has a small memory for migration,so as to achieve the effect of alleviating the pressure on node resources and improving the efficiency of virtual machine integration.3.This thesis describes the virtual machine placement problem as a multi-objective optimization problem that minimizes the waste of physical machine resources and minimizes the balanced load of the data center.Through the analysis of the data center virtual machine placement problem,it proposes a multi-objective optimization problem based on MOPSO(Multi-Objective Particle Swarm Optimization)algorithm of virtual machine placement algorithm.Firstly,by redesigning the particle coding scheme,the particle swarm algorithm can be applied to discrete problem solving.Then,based on the Metropolis criterion,the particles are allowed to accept inferior solutions with a certain probability in the iterative process to solve the problem of premature convergence of the particle swarm,and the position update method of the particle swarm algorithm is adjusted through the cross mutation operation of the genetic algorithm The diversity of particle populations.Finally,the global optimal selection and the maintenance of the Pareto solution set are used to compare the correlation value to improve the efficiency of the population search and enhance the multi-objective optimization ability of the placement algorithm.