基于多目标优化的应激式交换机迁移策略

Triggered Switch Migration Strategy Based on Multi-Objective Optimization

针对软件定义网络中控制器发生故障后,受影响交换机如何选取迁移目标的问题,综合考虑距离最近迁移和负载最低迁移的弊端,将控制器负载通过合理的建模转化为时延指标,结合控制器和交换机通信时延建立数学优化模型,设计遗传算法,基于全局最优的角度选择合理的迁移目标。仿真结果表明,该方法与最近迁移策略和利用率最低迁移策略相比,在整体网络负载较高的情况下,减少21%~40%新流响应时间。

To address the problem of migration target selection by the affected switch after the controller fails in the Software-Defined Network,this paper considers the shortcomings of the nearest migration and the minimum migration,and converts the controller load into a delay index through reasonable modeling.The controller and the switch communication delay are used to establish a mathematical optimization model,and the genetic algorithm is designed to select a reasonable migration target based on the global optimal angle.The simulation results show that compared with the recent migration strategy and the lowest utilization migration strategy,this method reduces the response time of new flows by 21%~40%when the overall network load is high.