Research On Availability-Oriented Virtual Machine Resources Management Technology

The heterogeneous and dynamics of cloud computing infrastructure, the dynamic join and exit of virtual resources, the software and hardware malfunction, operation maintenance, safety protection, and system upgrade may lead to the non-availability of cloud environment resources, bring great loss to the cloud computing service providers and cause users can not normally use the virtual machine and the use of cloud services. In recent years, by tracking and monitoring the availability of cloud computing resources in practical application, the researchers analyzed the influence of the usability of the virtual machine on the application performance. The results show that, if virtual machine resources maintain a higher frequency of non-availability of the task, the task execution, application performance and user QoS guarantee will be negatively affected. In the case of cloud task scheduling and resource allocation considering the availability of virtual machine, the overall performance of the cloud computing system will be significantly improved. Therefore, in order to research availability-oriented virtual machine resources management technology, the work is based on the virtual machine resource availability capacity and the prediction of the availability.First, in most existing researches, virtual machine availability is described as the probability of the virtual machine providing services in a period of time, while the problem that virtual machine is not available caused by the shortage of resource was ignored. To solve this problem, a method of computing virtual machine availability based on the failure distribution and the evaluation method of the virtual machine’s available capacity based on Markov-chain model is proposed. The method describes the probability of the virtual machine proposing service and the impact on the application services of the lack of resources in the future and this provides a basis for resource adjustment of virtual machine. Further, a new dynamic adjustment strategy of virtual resources (ADAR) is proposed. Simulation results indicate that, compared with the existing resource adjustment strategies, ADAR can make dynamic resource adjustment in time according to the availability of the virtual machine and effectively improve the utilization rate of system and guarantee the quality of service.Then, on the basis of the measurement method of the virtual machine’s availability, the thesis studies the forecast model for the value of virtual machine availability, and analyzes the applicability and accuracy of existing model. Grey-index combination forecast model, which is suitable to the changing virtual machine availability and this prediction model has high precision, is presented. According to the forecast model virtual machine availability, the virtual machine placement strategy based on prediction model is proposed. The effectiveness of the prediction model and the virtual machine placement strategy isverified through the experiments. Compared with the existing prediction model, the prediction accuracy of Grey-index combination forecast is higher.The thesis analyzes the related work of the virtual machine availability evaluation and resource adjustment, and describes a new availability evaluation of virtual machine. A strategy of the virtual machine’s resources dynamically adjustment and the virtual machine placement based on the prediction model of availability are proposed. To a certain extent, using this strategy can improve the utilization of the system, protect the service quality and reduce energy consumption.