A Study Of The Resource Self-Adaptive Allocation Management Technology In The Virtualized Environment

With the rapid development of IT services industry, service businesses and requirements are constantly changing. Service providers need to adopt effective resource management strategies to reduce operating costs. Virtualization technology provides a good solution for the realization of this strategy. By deviding the physical machine into many virtual containers, it provides good independence, security, and flexibility to dynamically deploy the shared hardware resources for the deployment of multi-service applications. Virtualized data centers consider the actual resources consumed by the server, and allocate resources among these containers in order to achieve optimal allocation of resources according to a particular cost-effectiveness scheduling settlement. Currently, most business service applications are concerned more about the quality of its service (QoS) and resource utilization. Data center need to provide adequate resources for every server application in order to ensure service performance, as while as dynamic adjustment of resource allocation, and avoid unnecessary waste.This paper refers to the virtual ization technology, service performance evaluation technology, and resource self-adaptive technology, and elaborates the principles and methods of performance evaluation, which can be used to make a quantitative analysis for the resource demands of server applications and help to develop resource allocation strategies. It also elaborates the self-adaptive model, the realization methods and current adaptive technology research in the dynamic resource allocation. In this paper, the Xen virtualization environment is used to deploy TPC-W e-commerce server applications, and the advantages of performance evaluation and resource management advantages through virtualization is also analysed. Fuzzy theory and game theroy are introduced to realize the self-adaptive control of resources.(1) Automatic control theory and technology are introduced into virtual resource management system. Fuzzy control theory is used to predict and control the system's resource requirtments. Fuzzy controller is running in a virtual server to perdict service physical resource needed for applications and periodically sends resource request message to the data center. (2) An improved algorithm of rules updating and conflict managing is designed, in order to achieve self-learning function of the rule base, and it also prevents the expansion of the rule base space and successfully avoids the rule conflict.(3) Game theory is introduced to resource management and a cooperative game model is designed to solve the competition between web server and database server under the conditions of limited resources to address. A utility function with parameters of throughput, average response time, etc. is picked. The existence and uniqueness of the convergence and equilibrium is demonstrated. There are also some experiments to verify the optimality of game theorem.(4) A performance evaluation program for e-commerce services is proposed, and according to the program, a user-feature based workload generator and performance data collector are designed. Testers can define the request type and workload strength according to their own requirements, by setting the workload parameters.In this paper, the system response time of server application is picked as the performance index to constraints the system CPU resource allocation and management. The result shows the self-adaptive model is a feasible solution in application environment, by comparing the performance datas and system status information collected in the two different resource allocation strategies. One is fixed value and the other is useing fuzzy control model. The experiments also show that the fuzzy controller has good control effect, not only ensured the quality of service, but also improved resource utilization to a certain extent.