Routing improvement to control congestion in software defined networks by using distributed controllers

One of the big challenges in software defined networks (SDN) is to find appropriate locations for controllers to shorten the latency between controllers and switches in wide area networks (WAN). In the literature, the majority of approaches are focused on the reduction of latency, but latency is only one of the factors of the overall cost between controllers and their associated switches. In this paper, we explore and investigate more possible factors of the cost, including the links utilization. In order to decrease the end-to-end cost, the concept of network partitioning is introduced and an Enhanced Clustering-based Network Partitioning Algorithm (ECNPA) is then proposed to partition the network. The proposed algorithm can guarantee that each partition is able to shorten the maximum end-to-end cost between controllers and their associated switches and improve routing by calculating bottleneck links. Extensive simulations are conducted under some real network topologies from the Internet Topology Zoo. The simulation results show that in the case of a busy network and the probability of congestion in it, the proposed algorithm has been able to well control congestion in the network by identifying the bottleneck links in each node's communication paths with other nodes. As a result, by taking into account the two factors of delay and the rate of busy links, the process of placement and distribution of controlers has been done with higher accuracy, and reducing the average of maximum end-to-end cost between controllers and their associated switches in Chinanet topology of China, Uunet topology of USA, DFN topology of Germany, and Rediris topology of Spain to 41.4694, 29.2853, 21.3805, and 46.4829 percent, respectively.