Improvement And Realization Of Virtual Machine Real - Time Migration Based On

Server virtualization, as the foundamental support technique in cloud computing datacenter, has many advantages, such as resource integration, live migration, rapidly clone and deploy virtual machines(VM). In this paper, we deeply research current server virtualization and live migration technique, Xen and other popular virtualization platforms. As one of the most important techniques, live migration of virtual machine can successfully migrate VM from one physical machine(PM) to another with only slight influence to upper business overload, and can be used in resource integration, bad balancing, power saving, fault-tolerance and machine routine maintenance.In spite of the advantages, current live migration technique is still not good enough in either efficiency or automation characteristic. This paper presents a novel approach Memory Exploration and Encoding (ME2) based on Xen. ME2first jumps into VM and differentiates pages useful depending on the OS memory management mechanism, and then utilizes Run Length Encode (RLE) algorithm to quickly compress the useful pages in order to efficiently decrease the total transferred data and total migration time. Experiments demonstrate that ME2can significantly decrease50.5%of total transferred data,48.2%of total time and47.6%of disruption time on average compared with Xen migration system.According to the good performance of ME2, we utilize such design philosophy in VM disk migration, and we demonstrate the disk exploration in details. Besides of the improvement of live migration itself, we designed and implemented an auto-migration prototype system, which can discover the hot-spot machine and automatically recognize the scenario pattern--load balancing or power saving, and then migrate the virtual machine from hot-spot to specified cold-spot physical machine. The process of migration can execute automatically and improve the efficiency of management.At last we overview the problems of the server virtualization and live migration and give an analysis of the technique in the future.