Research On Load Balancing Of SDN Controller Cluster

By decoupling the control plane and forwarding plane, software defined network provides great programmability for the dynamic network configuration and network control and shows great potential as an emerging network architecture. However, the traditional logical centralized control plane is usually handled by one computer. As a result, in a large scale network environment, like WAN, the traditional control plane inevitably affects the performance and scalability of the network. To overcome these problems, a new generation of clustering-control has emerged. By using super node and cluster shared area the clustering controller has enabled the information exchange between cluster controllers. But at the same time, these cluster controllers raise another problem that the loading imbalance between each controller in the controller cluster. The thesis probes into the problem and tries to provide the solution.In the first place, we point out the shortcomings of the traditional network architecture.Then we introduce the disadvantage of SDN when deployed in a large scale network environment. By introducing the cluster controller of the SDN, we describe and analyze the new generation controller. After that, we propose a method and a process to remapping the data forwarding plane to the control plane. Further, to meet the forwarding request of different network environment we propose two loading balance algorithms, Shared Pool Switch Schedule(SPSS) and Cost First Switch Schedule.The former balances the load by the rules of averaging the load of each controller, while the latter adjusts the mapping between the switch and the controller only when the adjustment can improve the performance of the cluster significantly.Finally, in order to verify the above algorithm we design the experimental scenes. The results show that SPSS and CFSS can improve the Load Balancing in SDN cluster controller.