基于改进遗传算法的SDN多控制器负载均衡机制研究

为解决软件定义网络中多控制器负载失衡问题，提出了一种基于非合作博弈降载的主控制器重选模型。首先，利用动态阈值来判别过载控制器；其次，采用基于优先权的迁移交换机决策机制；最后，构建以控制器集群的负载均衡度、平均总时延和交换机迁移成本作为效用函数的优化模型，采用改进的遗传算法求解，加入相似算子提高寻求全局最优解的速度及准确度。实验结果表明，该机制有效地均衡了控制平面的负载并优化了网络性能。

SDN multi-controller load balancing mechanism based on improved genetic algorithm

This paper proposed the non-cooperative game theory-based of load-reduction master controller reselection model in software defined networking to address the multi-controller load imbalance problem, called GTMCR. Firstly, the discriminating overload controller utilized the overload dynamic thresholds approach(ODT). Next, the migration of switches followed a priority decision switching strategy(PDSS). Finally, the constructed optimization model took the load balancing degree of controller clusters, average total delay and switch migration cost as utility functions. To prevent the global optimal solution from slipping into the local optimum, the solution value of optimal deployment used the genetic algorithm for improved multi-objective optimization(GAIMO) with the addition of a similarity operator to boost the algorithm''s convergence speed and the accuracy of the global optimal solution. The experimental results show that this mechanism effectively balances the control plane load while achieving the goal of optimizing network performance.