Research Of Multi-controller Load Balancing Technology Based On SDN

Software Defined Network(SDN),as a novel network architecture,solves the problems of poor business adaptability,insufficient flexibility and low deployment efficiency of the traditional network.However,limited by the controller's processing capacity and single point of failure,a single and centralized controller has become the bottleneck of the entire SDN network.Therefore,researchers have proposed a distributed control plane composed of multiple controllers.But at the same time,due to the time-varying and mutability of network traffic,the distributed control plane causes load imbalance among multi-controllers,resulting in degraded network system performance and failure to maximize controller performance.Based on the above background,this thesis focuses on the load balancing problem of multi-controllers in the master/slave coexistence mode and the equal coexistence mode.The main research contents of this thesis are as follows:(1)We propose a dynamic load balancing algorithm based on Efficiency-Aware Switch Migration(DLB-EASM)for multi-controllers.Firstly,the load of controller is quantified from four aspects: switch request,network domain management,network status synchronization,and flow table installation.Then,for the problem of rigorous selection of migration objects leads to low load balancing efficiency in traditional load balancing algorithms based on switch migration,we establish the efficiency model of switch migration by combining the system overhead caused by the switch migration process and the impact on the network performance after the migration,and design a switch migration objects selection strategy based on this efficiency model.Finally,the remapping between the controller and the switch is accomplished through a four-phase migration protocol.Experimental results show that DLB-EASM can effectively achieve load balancing for multiple controllers,reduce the response time of controller,and improve the throughput of the system.(2)We propose an adaptive traffic scheduling algorithm based on NonCooperative Game Theory(ATS-NCGT).In the equal coexistence mode of multicontrollers,we model the load balancing problem of multi-controllers as a non-cooperative game between switches in the data plane based on the interaction characteristics between the switch and the controller.Then,in order to improve the optimization accuracy and convergence speed in the solving process of the load balancing game,we propose a hybrid genetic algorithm based on simulated annealing,and use the switch's expected response time as the payoff function to find the optimal solution for traffic distribution strategy.Based on the above methods,we implement a non-cooperative game-based traffic scheduling algorithm to achieve the load balancing of multi-controllers.Experimental results show that compared with other similar algorithms,ATS-NCGT can distribute traffic more evenly and obtain a higher load balance rate among multi-controllers.In terms of controller response time and controller resource utilization,ATS-NCGT has a better comprehensive performance.