Research On Virtual Machine Deployment Strategy In Cloud Environment

With the continuous progress of resource integration process,the development momentum of cloud computing is only increasing without decreasing in recent years,more and more enterprises choose to deploy applications on the cloud platform.However,it is a challenge for cloud platform to manage the virtual machine whose scale continues to expand,and how to solve the virtual machine deployment problem is the current research hotspot and difficult problem.Based on the review of many articles and summarizing the research results in relevant fields,this paper proposes a Multi-Objective Particle Swarm Optimization algorithm combining Mutation Operator(MOPSOM)with the goal of improving resource utilization,reducing response time,and reducing the costs of task scheduling,shortening the completion time,but also on the issue of virtual machine deployment made the following series of preparations.First of all,the related technology are summarized.In the task and virtual machine mapping phase,according to the data storage requirement and task dependent feature,constructing cost model and time model of task,and modeling task scheduling sequence;in the virtual machine and the server mapping stage,according to data center server energy consumption target and user level agreement,build the server the resource utilization model and the response time model of the system,so as to lay foundation of the virtual machine deployment phase.Secondly,this paper proposes a virtual machine deployment scheme that bases on particle swarm optimization and combines with genetic algorithm mutation operator.We consider the mapping problem that this paper wants to solve by analyzing the key points of initial allocation and the frequency of migration completely.The core idea is to control the response time of the server while improving the resource utilization,so as to avoid violating the user level agreement.The benefits of this approach not only reduce the waste rate of resources,but also reduce the number of physical machines which is running,thus reducing the energy consumption of data center.In addition,controlling the response time can effectively avoid the long response time,resulting in the decline of customer service quality.More importantly,controlling the response time length promotes server load balancing,thereby reducing the number of times that the virtual machine needs to migrate due to server overloading during the running process.The simulation results show that the deployment strategy can achieve the optimization goals,and the convergence is also better.Finally,this paper proposes a task scheduling method that combines particle swarm optimization with genetic algorithm mutation operator.The task scheduling problem is solved by considering the task dependent data storage problem,the execution cost and communication cost of the task,the user is more concerned about completion time.The core idea is to determine an execution sequence of the task according to the scheduling model,and then obtain the scheduling scheme that makes the cost and the makespan as low as possible according to the objective function defined by the particle swarm.Due to the contradiction between task scheduling cost and completion time,therefore,this paper chooses the method of weight ratio to deal with multi-objective optimization problems.So,the optimal task scheduling scheme is the point where the particle is found in the search space,which makes the minimum of the objective function.Simulation results show that the proposed algorithm can effectively shorten the completion time of the task,and from the scheduling results,the more dependent tasks and data are basically assigned to the same node.