An Intelligent Congestion Control Method in Software Defined Networks
Jihong Zhao, Mengfei Tong, Hua Qu, Jianlong Zhao · 2019
Congestion control tends to be a quite significant problem with the increment of internet traffic. Conventional networks treat packets drop as an indication of congestion and employ the end devices to mitigate congestion. This may impose restrictions on network performance with the absence of comprehensive view of network. To solve these problems, this paper proposes an intelligent congestion control algorithm based on the global characteristics of software defined network architecture. The controller monitors the global operating state of network and evaluates instant congestion quality of backup paths in network by fuzzy logic. Combing with reinforcement learning, the controller evaluates the rerouting decisions under current network circumstances and chooses the optimal forwarding path which meets the traffic load requirement to forward flows. Experiments shows that the mechanism can significantly enhance network performance especially when network is under heavy load.