A proactive controller failure recovery mechanism in SD-WAN with multiple controllers

Abstract

The separation of data and control plane is a key feature of Softwaredefined networking (SDN), which makes network administration more intelligent. The control plane is realized using a logically centralized controller. In SD-WAN, as the network becomes larger, multiple controllers are needed to manage the network. In this scenario, there is a chance for controller failure due to overload. Once the controller fails, the switches lose connection with the controller. The load of the failed controller is to be re-distributed among other controllers. Sometimes this load transfer may cause many other problems like controller chain failure. It may consistently undermine the trustworthiness of the network. In the proposed technique, the multiple controller deployment based on affinity propagation clustering in SD-WAN is extended to include a proactive controller failure recovery mechanism. Whenever the controller load exceeds a pre-defined threshold, a set of switches under the bottleneck controller will be reassigned to a neighbouring controller without exceeding its capacity. The simulation results show that when network traffic increased, the proposed proactive controller failure recovery method balanced the controller load, resolved a cascading controller failure, improved the average throughput, and reduced the average end-to-end delay and packet loss effectively.