Research On Applications Migration Strategy And Resources Management Technology Towards Cloud Environment

Cloud computing with its dynamic scalability, cost-effective, on-demand features to provide reliable service, becomes the focus of industry and academia in recent years. More and more small and medium enterprises and individuals want to move their applications to cloud platforms, meanwhile, there is lack of a universal, simple and effective framework and strategy for application migration, and reasonable benifits estimation under this policy approach, the resources scheduling method under the application migration strategy and certain conditions also need to be reconsidered. (By the analysis of series issues above, the main contents of this thesis are as fallows) Through a series of questions on this analysis, the main contents of this thesis are as follows.The proposed application migration strategy take a server as a unit, the application of the server as the second floor, part of the application migrate, part of the application keep to the local, as a hybrid migration model. Using the weighted undirected graph establish relationship of traffic between servers, convert the solution of figuring the collection of intended migration server to the graph K-partitioning problem, the goal is to minimizing the data-flow between the intend migration part and the keeping part. A particle swarm optimization algorithm with genetic algorithm is proposed to solve the integration of K-partitioning of the NP computational problems, trying to minimize the cost of migration and to maximize benefits, and to determine the conditions which help us to decide whether migrate application servers and the corresponding collection. Experimental results show that the proposed algorithm compared with the traditional PSO algorithm has greatly improved the accuracy with faster computing to find the minimum flow divided.We estimate the benfits of migration from the perspective of dynamic resource allocation, by considering the features of cloud such as dynamics resources, the characteristics of distribution according to need, we proposed a linear and Winter seasonal exponential smoothing forecasting method resources on the migration of some applications using dynamic resource allocation to address the response to the cyclical peak traffic, improve service quality, cost-saving use of resources and other issues. Using commercial operation data to analyze the method, experimental results show that taking the dynamic method to configuration the resources of migrated application can increase 43% of revenue for the users in a year.Because the resource scheduling algorithms under the application migration conditions need to take into account the conditions of the application migration as a whole, and reconsider dependence relationship between the applications, in this case, the cloud computing service providers under this specific scenario has to consider more complex situations. This paper presents an improved genetic algorithm for resource scheduling algorithms aim to improve the efficiency and resource utilization, and ensure the user quality while reducing as much as possible the cost of cloud computing service providers. Experimental results show that the proposed resource allocation algorithm in comparison with some traditional algorithms has better algorithm performance and resource utilization.