Research On The Method Of Dynamic Deployment Of Virtual Machine Based On Prediction Of CPU Utilization

In the cloud environment, because of the change of the system condition and the dynamic changes of the applied load, the work to ensure the quality of cloud services has brought new challenges. With the continuous change of virtual machine of the demand for resources, there will be a lot of idle resources in parts of the server, and a serious shortage of resources in other parts of the server, the system leads to uneven distribution of resources, resource utilization fell, severely affecting the quality of services. So in the face of complex services and applications, under the premise of ensuring the quality of service, how to improve the resource utilization of the system, is one of the focuses of current research for service computing.Based on the research of traditional virtual machine deployment method, the research on the method of dynamic deployment of virtual machine based on prediction of CPU utilization in the cloud environment was proposed in this thesis. It allocated the virtual machine instances dynamically and generated the reasonable dynamic deployment plans real-time and effectively according to the CPU utilization of each server in the cloud, to achieve multi-goals such as the high resource utilization, low cost of services, high user experience and service quality. In this thesis, the research work mainly include:first, research on the process of dynamic deployment of virtual machine based on prediction of CPU utilization, analyses the insufficient in the study of traditional virtual machine deployment, defined the main research contents and key problems in the process of dynamic deployment; Second, CPU utilization prediction algorithm oriented to dynamic deployment of virtual machine was proposed, first of all, using ARIMA model to predict the CPU utilization time sequence of linear main body of the server, and then use BP neural network model to estimates the nonlinear residual and eventually superposition for the entire time series prediction results; Third, research on the method for generating dynamic deployment scheme of virtual machine based on the CPU utilization forecast. Target for virtual machine dynamic deployment problem, establishes the optimization model, and then separately from the timing of the virtual machine dynamic deployment, the choice of being deployed virtual machine and purpose server option three aspects on the problem of dynamic virtual machine deployment the description of the specific analysis, Finally, the feasibility and effectiveness of the proposed method has been verified through the experiments. Experimental results prove that the feasibility of the prediction algorithm of the service performance and the generation method of deployment oriented to virtual machine dynamic deployment and the certain effectiveness and accuracy.Finally, we design simulation experiments to prove the accuracy of the algorithm oriented to virtual machine dynamic deployment of CPU utilization prediction, and the effectiveness of the proposed method for generating dynamic deployment scheme of virtual machine based on CPU utilization forecast. Make the system to meet the goals of SLA, improve the utilization of resources at the same time. Therefore, the method of dynamic deployment of virtual machine based on prediction of CPU utilization in the cloud environment is an important reference for subsequent research.