Research On Virtual Machine Resources Dynamic Allocation Method Based On Revenue In Cloud Computing

With the development of distributed computing, parallel computing and grid computing, cloud computing begins to form and continues to improve. Cloud computing is based on virtualization technology, using virtualization technology, IT resources form into a large resources pool. And the resources in pool are as a service for client use. Virtualization technology can reduce the number of physic servers in the cloud data center or large scale resource pool by server consolidation to reduce power and extra cost consumption, increase server reliability and significantly simplify the IT infrastructure, optimize the resources and reduce risk. It also brings challenges to the optimization of resource allocation method due to the changes in the underlying architecture and the continuing changes of the requirement of services in the virtual machine. As the demand for resources of the client service running on the virtual machine, how to allocate resource to the virtual machine effectively and dynamically in the resource pool to improve the utilization of the computing resource and to meet the performance requirement of the client service that running on the virtual machine is an important research subject in the resource allocation of virtualized computing environment.This thesis conducts a wide range of theoretical and technological research which focuses on cloud computing based on virtualization. By studying, I mastered the virtualization cloud computing technology. In second chapter, elaborating the classification of the virtual machine resource allocation method in cloud environment and classification of performance prediction model. After that, detailed expose of based on the revenue of the cloud environment virtual machine resource allocation process, and aiming to discuss and research the performance of service on the virtual machine and the virtual machine resource allocation process based-on revenue. The main contents include:First, considering the revenue issue of cloud resources provider and the performance of the service on the virtual machine problem, a dynamic allocation process of the virtual machine in cloud environment based on revenue is presented. Also, the dynamic resource allocation architecture is given.Second, for the performance of service running on the virtual machine, performance data is processed by data preprocessing method. And stochastic gradient regression model is used to predict the average response time of service which provides the time for resources dynamic allocation unit when to resources allocation.Third, due to the resources demand of service running on virtual machine is unknown, and the gray prediction method is adopted to predict the pre-demand of resources.Fourth, according to the user’s budget and the average response time, the cloud divides the level of service which runs on the virtual machine. In accordance with the principle of maximizing the revenue, an economic revenue-based dynamic resource allocation method is proposed which satisfies the client requirement. So the cloud resources provider obtains distinct revenue and the method establishes the relationship between resources number and revenue.In this thesis, the simulation is executed to verify the stochastic gradient regression algorithm to predict the service average response time, and compares to the SVM algorithm to illustrate the stochastic gradient regression algorithm having higher accuracy. For the resource allocation method, the thesis adopts basis cloud computing platform hadoop as an example of requirement to verify the resources allocation method is effective and feasible.