Research On Controller Placement Problem Based On SDN

With the updating of communication technology,SDN network emerges as the time require,and the controller is the core of the SDN network,which is responsible for the control of the whole network.For large networks,the management of a single controller cannot meet the needs of all switches.It is necessary to use multiple controllers to share the communication pressure of the whole system.The communication delay between controller and switch plays an important role in the performance of the whole network.So the reasonable deployment of the controller location can effectively improve the performance of the whole network.In this paper,a Revised K-means based Controller Placement Algorithm in Software Defined Network,RKCP is proposed.The algorithm consists of three parts: 1.Through the revised K-means algorithm,the network is preliminarily clustered,and the minimization of average-control latency is achieved.2.Through the node annealing algorithm to select the boundary nodes and the idea of simulated annealing algorithm is used to reallocate the boundary nodes to achieve the load balancing among the controllers.3.Through the Tag delivery algorithm to find isolated nodes and reallocate them,the problem of cross-domain communication is solved.Simulation results show that the proposed algorithm can minimize the average control latency and the controller load balance index when there is no isolated nodes.In view of the characteristics of multi-controller placement problem,this paper proposes a Bat Inspired Controller Placement Algorithm in Software Defined Network.The algorithm combines the BAT algorithm with the multi-controller placement problem.Through the Tag delivery algorithm to remove isolated nodes.Achieving the minimization of average-control latency by narrowing the average-control latency.Achieving load balancing among the controllers by narrowing the load utilization.Simulation results show that the proposed algorithm can minimize the average control latency and the controller load balance index when it meets the load utilization limit and there is no isolated nodes.