The Research Of Migration Strategy In Virtual Environments

The Data Center provides powerful computing capabilities, but in practice much of this power goes to waste, due to static resource allocation strategies. Service consolidation through employing virtualization technology can solve such problems in some degree. Moreover, most of the data center running the Web application, which has properties of dynamic load, easy to bring transient spikes and flash crowds. Migration of virtual machines can improves the utilization ratio of resource and the balance of load. However the virtual machine's migration is mainly operated by manual currently, the reaction is slow which leads to prone to error, and lack the dynamic control. How to migrate virtual machine automaticly according to the characteristics of web application is one of the hot topics in research community.At present, the automatic migration system mainly focuses on researches the problem of hotpot detection, resource allocation and migration strategy and so on. Based on researching and analyzing the above issues and related technology. Due to their disadvantages, we propose the detection standards of hotspot, and two migration strategies based on different objectives. Furthermore we apply our approach to a real environment, and make a comparison of the migration strategies through simulation to evaluate their performance. Overall, the main work we done is:(1) We observe and track the three important resources in the data center, in order to adequately address the issue of hotspot detection, we collect the resource utilization, using the time series analysis methods, predict the trends of the use of resource utilization, judge the hotspot with the mathematical expectations of resource utilization and the predicted values.(2) We study Black-Gray box migration strategy. Whose problem is: the migration success rate is too low, migration times are too high, and the workload is imbalance after the elimination of hotspots. Then we propose heuristic migration algorithm called Balance algorithm based on the permutation and combination and BFD heuristic algorithm, the algorithm focus reducing the migration times, improving the migration success rate and workload balance after elimination of hotspots. According to different objectives, we also propose Cost algorithm which focus on reducing the number of physical node, improving the migration success rate and workload balancing and system stability after elimination of hotspots.(3) We implement the automatic migration system, which can automatic migrate according to dynamic load. We simulate the two migration algorithm and black-gray algorithm, analyze and compare them through simulation experiment, which verify it is an effective and reasonable, and improved the performance of physical nodes in each data center.