Research On Load Balancing Strategy Of Multiple SDN Controllers Based On Deployment And Migration Optimization

With the widespread application of Software Defined Networking(SDN)in 5G terrestrial networks and satellite networks,the SDN control plane has evolved from a single-controller architecture to a distributed multi-controller architecture.Although the distributed controller improves the processing capability of the control plane,it introduces new problems,that is,where the controller should be deployed and how to balance the load among the controllers,which has become one of the current research hotspots.In this paper,the controller deployment optimization problem is established based on the flow setup delay.By analyzing the influence of cross-domain traffic size on the flow setup delay,we transform the optimization problem into a graph partitioning problem.This paper proposes an improved Metis algorithm to divide controller domains,which can significantly reduce cross-domain traffic and optimize flow setup delay under the premise that controller loads are basically balanced.After division,this paper deploys a controller in each control domain based on the intra-domain weighted delay.When the satellite network topology and satellite load change,the original deployment is no longer applicable,and the controller needs to be redeployed.In this paper,the ParMetis algorithm is used to re-divide the control domain and deploy the controller based on the division result of the previous time slot.The algorithm fully considers the original division,optimizes the size of cross-domain traffic and reduces switch migration costs.The simulation results show that the improved Metis algorithm and ParMetis algorithm are superior to the existing schemes in terms of total controller load,flow setup delay and other indicators.In addition,this paper studies the impact of flow rule deployment on controller load,and establishes rule deployment and switch migration optimization problems based on end-to-end delay.According to the feature that wildcard rules can match more flows,this paper proposes a CostAware Routing algorithm(CAR).When routing traffic,the CAR algorithm makes full use of existing rules and preferentially installs wildcard rules to improve flow table utilization and reduce controller load.Furthermore,considering the cost of switch migration,this paper proposes a BenefitCost Switch Migration algorithm(BCSM)to maximize the switch migration benefit.The simulation results show that the CAR algorithm significantly reduces and initially balances the controller load,and combined with the BCSM algorithm,it can reduce the end-to-end delay at a small migration cost.