Research On SDN-based Multi-controller Load Balancing Technology

SDN,Software Defined Network,is an implementation of network virtualization.Its core technology,OpenFlow,separates the control plane and data plane on the network device and shields the differences between the underlying network devices.The centrally managed controller implements intelligent control of traffic by running user-defined network routing and transport rules.At the same time,when the centralized control plane responds to a large-scale network environment,it will cause bottlenecks in network system performance.To this end,multi-controller cluster technology has been proposed,but at the same time,the mapping problem between the switch and the controller is also derived.That is,the static configuration of the switch and the controller cannot adapt to the dynamic change of the network flow,thereby causing the load between the controllers not to be Equilibrium makes the controller cluster a bottleneck in the performance of the entire network.In this thesis,the multi-controller load balancing of control plane in SDN is studied thoroughly.The advantages of SDN network,the performance and scalability of largescale deployment are analyzed.Switch remapping and load balancing of controller cluster are proposed.The contributions of this thesis are as follows.The problem of switch remapping when controller is overload is studied.The principle of stability and self-interest maximization are used to model the target controller selection problem in the remapping process of the SDN controller cluster into pure strategic Nash equilibrium of non-cooperative game.Based on switch remapping,a load shedding algorithm is proposed.The experimental results show that the proposed algorithm has significant effects on the load shedding when the controller is overloaded,and finally the response time can be reduced.The load balancing problem between multi-controllers based on master-slave structure in SDN controller cluster is studied.based on genetic and simulated annealing algorithms,a genetic annealing algorithm is designed to solve the problem of packet-in event allocation.It combines the advantages of fast convergence of genetic algorithm and avoids failing into local optimal solution of simulated annealing.The experimental results show that the proposed packet-in event allocation scheme has a good effect on load balancing,and finally achieves the effect of reducing response time.