Research On Dynamic Load Balancing Of Data Center Network Based On SDN

With the development of Internet technology becoming closely with people’s daily life, the demands for the Internet technology gradually increased. In recent years, cloud computing technology is developing rapidly, and then data center to be established in the world. In the environment of various demands for cloud services widely arisen, the traffic pressure of a data center network faced will continue to increase.How to make the best use of limited net resource has become one of the most important problem need to be solved. The development of network technology is limited by the defects of TCP/IP architecture for traditional network, the low utilization rate of the network bandwidth is one of the common existing problems, and these problems have hindered the development of load balance technology in traditional networks.The concept of software defined network completely subverts the architecture of traditional network, and provides ideas to solve this problem. We can obtain real-time network status information in the controller through openflow technology, How to manage the network and solve the problem of load balance in traditional network will become easier.The main research work of this thesis are as follows:1) Explain the concepts and technical knowledge, including openflow technology based on SDN, data center network topological structure scheme and load balance technology in traditional network. the research focuses on the overall architecture of openflow technology and the principle of the openflow switches and the existing openflow controller.2) Establish a fat-tree network model based on data center,introduce the algorithm design process of openflow network dynamic schedule load balancing of DSLB. Expounds the overall structure of Ryu controller, module functions and development methods, and the realization of DSLB algorithm in the Ryu controller.3) Program for fat-tree topology under the Mininet simulation platform. Experiments in different environments with connect to the controller. Compare the experiments’ results of two different algorithm according to the load balance performance evaluation index and verify the validity of the work proposed DSLB algorithm.