为了确定控制器的最优化部署方案,构建软件定义网络中逻辑上集中、物理上分布的控制平面,提出软件定义网络中应用二值粒子群优化的控制器部署策略。对控制器部署问题建模,以交换机到控制器的平均时延最短以及在网络中部署的控制器数量较少为多优化目标。提出粒子重构机制,实现粒子群优化算法的二值化,用以表示控制器在网络中部署的位置。基于二值粒子群优化算法设计多优化目标的控制器部署策略,仿真得到控制器部署问题的非劣最优解集合,对应给定的控制器数量,得到平均时延最小的控制器部署方案。实验结果表明,应用二值粒子群优化的控制器部署策略联合考虑了控制器数量和交换机到控制器的平均时延,为实现控制器最优化部署提供了依据。
A strategy of controller placement in software defined networks using the binary particle swarm optimization is proposed to find the optimal placement scheme of controllers and to build a logically centralized and physically distributed control plan for the controller placement problem in SDN.An optimization model is built for the problem,and the optimization goals are to minimize both the number of controllers and the latency from controllers to switch.The binary particle reconstruction mechanism is put forward to denote the position of the controller in the network.A binary particle swarm algorithm is applied to find the set of Pareto optimum for the problem,from which a strategy of controller placement is provided.Simulation results show that the proposed algorithm gets the optimal placement strategies with minimum latency for different number of controllers.