Abstract: In multi-controller Software Defined Network (SDN), since the existed switch migration strategies can not adapt to the change of switch traffic which only consider single migration factor, a Switch Migration Strategy based on Improved Gravitation Search Algorithm (IGS-SMS) was proposed. In the decision-making stage, the multi-objective decision based on fuzzy satisfaction was used to optimize the objectives by the competitive priority of membership. In the calculating phase, the objective function with the top priority was optimized by improved gravitational search algorithm. The simulation results show that the IGS-SMS achieves good load balancing of controllers while ensuring the index of transmission delay and switch redistribution. When local load was heavy in the experiment, Dynamic Switches Migration Algorithm (DSMA) and Progressive Auction based Switches Migration Mechanism (PASMM) couldn't alleviate overload. By contrast, IGS-SMS could alleviate overload, and load balancing degree was lower than DSMA and PASMM.

Key words: Software Defined Network (SDN), multi-controller, switch migration strategy, Gravitation Search Algorithm (GSA), fuzzy satisfaction

摘要： 针对多控制器软件定义网络（SDN）中交换机迁移策略迁移代价衡量单一，不能适应交换机流量的变化的情况，提出基于改进引力搜索算法的交换机迁移策略（IGS-SMS）。在决策阶段，应用基于模糊满意度的多目标决策方法，优化目标根据隶属度大小竞争优先权；在计算阶段，通过改进引力搜索算法优化优先权高的目标函数。实验结果表明，IGS-SMS在实现负载均衡的同时，能保证传输时延与交换机重分配的指标；在实验中，当局部负载较重时，动态迁移算法（DSMA）和基于改进型拍卖交换机迁移机制（PASMM）不能缓解控制器过载，而IGS-SMS执行后无控制器过载，且负载均衡度小于DSMA和PASMM。

关键词: 软件定义网络, 多控制器, 交换机迁移策略, 引力搜索算法, 模糊满意度