Research Of Virtual Machine Incremental Migration In Cloud Computing Environment

As a hotspot of research and application in recent years, more and more ITcompanies and insiders focus on cloud computing, which is considered as the newdevelop model of next generation internet, will bring about tremendous changes topeople’s production and life. With cloud computing, network service providers canprocess millions or even hundreds of millions of data as soon as possible, as has theequally processing power of supercomputer. At the same time, end-users can demandthe use of these resources and services through network, and achieve the dream of usingresources as public facility, they can use cloud services as using water and electricity.Virtualization is one of the core components of the cloud computing system; itrealizes the logical abstraction of IT resources and unified representation, makes storageresources fully integrated and efficiently used, and provides a good solution for theeffective management and distribution in cloud computing system. Live migration ofVM technology is the most valuable application, and it’s the main content in this thesis.VM live migration is one of key technologies in virtualization technology. Livemigration means that a running VM migrates from one physical host to another physicalhost without stop, this technology has been widely used in terms of load balancing, faulttolerance, and energy management. There will be a short downtime during migration,users will not be significantly affected, but in the environment of real-time requirement,how to reduce negative impact to the users is particularly important. In this thesis, wechoose the open source virtual machine monitor Xen as research platform. The mainwork of this thesis as follows:1. Introduce the history and develop prospects of virtualization technology andcloud computing, and close relationship between cloud computing and virtualizationtechnology in terms of server virtualization, application virtualization, Etc.2. Analyze the live migration technology in terms of migration concept, migrationcontent, migration process and migration evaluation, and in-depth study the architectureof Xen, point out the inadequate performance in high-rate dirty page rate environment. 3. This thesis presents the optimized live migration framework, mainly added twomodules: the page statistics module and page layered module. The main work of pagestatistics is to record change history of memory page in Pre-copy period, to determinewritable working set (WWS), The main work of dirty page layered module is to layermemory pages according to statistics result, the few visited pages are transferred firstand the others are pushed back, the problem of dirty pages being migrated repeatly initerative process is resolved to a certain extent; effectively reduce migration time anddowntime.4. On the basis of study Xen live migration source code, realize the two modules.Proposed page statistics algorithms and page layered algorithms; this two algorithmswork together achieving the purpose of reducing total migration time and downtime.Finally, we verify the feasibility and effectiveness of the algorithms, prove that theoptimized live migration framework can effectively improve the performance ofmigration in high-rate dirty page environment.