Study On Resources Restructuring Of Virtualization System

Cloud computing is the product integration of utility computing, parallel computingand grid computing. It provides IT services including computing, storage and networkby pooling resources. However the expanding ultra-large-scale cloud computing datacenters have led to difficulties in managing and operating costs rising dilemma,particularly in: Resources utilization rate still not high, amazing energy consumption,unreasonable virtual machine resources allocation. Elastic cloud computing resourcemanagement features two major considerations: the user QoS and revenue of cloudservice provider. Most virtual machine users still schedule their VMs configuration bythe peak demand resource allocation of virtual machines, resulting in most periods ofidle resources and servers long time low-load running, which wastes a lot of power. Theacademic literature, aiming at resource utilization issues, have proposed resourceallocation and deployment method for virtual machine cluster, and the best resourcesconfiguration which meets both user QoS and the demand of not wasting too manyresources is closely related to load type and intensity upon the virtual machine. We,aiming at the inadequacies of the hypothesis in most literature about relations betweenindividual virtual machine’s optimal resources allocation, conduct in-depth study,meanwhile employ game theory to address problem of reorganization of resources tooptimize different resource-preferred virtual machines’ performance under conditions oflimited resources.The main work is summarized as follows:1)Based on virtual machine-web server scenario, we use queuing theory to modelthe average request rate, the average service time, resource average utilization, andaverage response time the web service request test scenario, and combined with theclassic formula of queuing theory Little relations to derive the relation among theaverage response time, average request rate and average service time; through the dataobserved in the process of web services running T time, including resource utilization,average response time and average request rate,we have conducted consistency checkover the model, and calculated the average service time for web request executed onvirtual machine; in eight groups’ experimental test controlling load intensity variable,we get the load intensity corresponding virtual machine resources critical point, when avirtual machine resource allocation comes below the critical point, the response time of which will sharply decline. In the final by regression analysis of the experimental datasignificant relationship between the rate and the best service request allocation ofresources is obtained.2) To the issue of restructuring the virtual machine resources under conditions oflimited resources, corresponding solutions, based on whether the virtual machine’s bestresources allocation decisions is conflict or not, were respectively proposed. Based onthe above performance computing model, the virtual machine depending on the loadtype and intensity independently compute and apply for the resources required to deploy,if the sum of the target virtual machine resource allocation of resources does not exceedthe initial sum of the restructuring, then according to their needs directly re-allocateresources, otherwise consider reorganization of resources conflict. For this problem,game negotiation is employed. Experimental results show the effectiveness of thismethod as the following points: improving resource utilization, optimizing virtualmachine performance, ensuring fairness about resource redistribution.