Application Research On Multiagent Genetic Algorithms For Resource Scheduling In Cloud Computing

At present, the cloud computing and virtualization technology are got in a rapid development. The main purpose of cloud computing is that it can provide a high-performance and high-reliability computing ability for all users. The resources scheduling is an important part of cloud system which can directly impact on the performance of the whole cloud computing system. The thesis firstly focuses on the study of resource scheduling theory and technology in cloud computing with diverse user requests, and then it does some researches on the resource organization types and the common used resource scheduling framework for the open cloud computing system.The basic theory and algorithm process of Genetic Algorithm (GA) and one of its hybrid algorithms-Multi-agent Genetic algorithm (MAGA) are studied in details. Then discussed in detail on the limitations of GA when it is used to solve optimization problems with high dimensional datasets. The experiments results show that MAG A has a higher convergence rate and better parallelism than common GA when it is used to solve high dimensional function optimization problems based on the optimization results and the average number of required iterations. And also, it was proved that MAGA is more suitable for the high-dimensional, large-scale dataset optimization.At last, designed a cloud computing simulation system based on virtualization technology, and the MAGA and the scheduling policy of group requested are used in the system for resource scheduling. In order to verify the feasibility of the proposed strategy, there set up three different types of experimental strategies for the testing of the MAGA and traditional strategies such as round-robin scheduling and Min-min scheduling. The testing results show that MAGA designed here can obtain a more favorable equilibrium result and better performances when dealing with heterogeneous and virtualized resources, because it has a more comprehensive consideration about the system’s overall condition. Therefore, the conclusion is that MAGA for resource scheduling of cloud computing is feasible and effective.