Considering a software defined network, distributed controller architecture uses multiple controllers in which each controller manages a part of the network. The load imbalance problem in this architecture causes a large number of switch migrations resulting in a significant increase in switch migration cost and average network response time along with a decrease in throughput. Although recent studies have addressed these issues, access to optimal response time had been achieved with high cost of switch migration and sometimes with reduction of throughput using their methods. Therefore, the load balance in the present study is managed by a variable threshold based on the controllers’ workload. In other words, migration is done by selecting optimal switch and controller so that the switch will be selected with the lowest traffic generation rate which could return the source controller to its steady state. Using the suggested method, a destination controller is selected based on some important parameters such as CPU utilization, rate of incoming packets and the number of hops between switch and controller. The TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) algorithm is used to select the best controller based on the above-mentioned criteria and the AHP (analytic hierarchy process) algorithm is employed for determining the ratio of each criterion. The proposed method considerably outperforms other methods by achieving about 6 and 78% improvement in throughput and the number of switch migration in our implementation, respectively.

中文翻译：

---

一种用于软件定义网络中交换机迁移的增强型基于 AHP-TOPSIS 的负载均衡算法

考虑到软件定义网络，分布式控制器架构使用多个控制器，其中每个控制器管理网络的一部分。这种架构中的负载不平衡问题导致大量交换机迁移，导致交换机迁移成本和平均网络响应时间显着增加，同时吞吐量下降。尽管最近的研究已经解决了这些问题，但通过交换机迁移的高成本和有时使用他们的方法降低吞吐量，已经实现了获得最佳响应时间。因此，本研究中的负载平衡由基于控制器工作负载的可变阈值管理。换句话说，迁移是通过选择最佳交换机和控制器来完成的，以便选择具有最低流量生成率的交换机，从而使源控制器恢复到其稳定状态。使用建议的方法，根据一些重要参数（例如 CPU 利用率、传入数据包的速率以及交换机和控制器之间的跳数）来选择目标控制器。TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution)算法基于上述准则选择最佳控制器，AHP(层次分析法)算法用于确定每个准则的比率。通过在我们的实现中实现吞吐量和交换机迁移数量的大约 6% 和 78% 的改进，所提出的方法大大优于其他方法，