Research On Strategy Of Controller Placement In Software Defined Network

Software Defined Network(SDN)as a novel network architecture realizes the centralized control of the network by decoupling the forwarding and controlling functions,which greatly improves the flexibility and controllability of the network.With the SDN gradually extended to large network such as wide area network,Using multiple controllers to construct distributed control layer is the inevitable choice for the application of SDN in large networks.However,it is still a problem that how to realize the reasonable placement of the distributed control layer in the SDN architecture network.As the primary task of control layer is to deal with network events in time,the delay of communication between the controller and the switch is regarded as an important index to evaluate the performance of control layer placement.During the operation of the network,it often occurs the failure of communication links and controller components.Among them,the links failure will directly affect the network topology,resulting in sharp deterioration in delay performance.When the controller fails,the network forwarding function will be affected,and even lead to local network paralysis.In view of the above problems,the placement strategy of control layer in the paper is designed from two aspects: the initial placement of the control layer and the replacement of the later controller.In the initial placement design of the control layer,the paper comprehensively considers the normal states of the network and the worst-case delay minimization problem of multiple links fault states,and the network state delay is used as new optimization goals to establish a control layer placement model based on the network state,and design two control layer placement algorithms based on greedy algorithm and particle swarm algorithm to solve the model.The simulation results show that the optimal number of controllers is three in the application scenario.At the same time,the two algorithms can guarantee that the delay index maintains in 10.7ms range under 74% and 67% of the network state.Aiming at the problem of controller failure in the process of network operation,we propose a scheme of control layer replacement based on switch migration mechanism.The scheme considers three factors of the delay overhead and load balancing overhead and cross-domain communication overhead,and constructs a control layer replacement model,and designs a controller replacement algorithm based on genetic algorithm.Simulation results show that the proposed algorithm avoids the cross-domain communication problem resulting from switch migration in the control layer,and effectively guarantees the overall performance of the control layer after the replacement.