Proactive Fault-Recovery on Distributed Software-Defined Network Controllers

Abstract Distributing the control plane on Software-Defined Networks (SDN) is a promising approach for improving network robustness. Using this approach, the control plane becomes physically distributed but logically centralized. This distributed SDN controller architecture reduces the controller's overload and enables a quick response time between controllers and switches. Besides, in case of network or controller failure, it decreases the time needed for fault recovery. In this work, we propose an approach to reduce the average recovery time on faulty scenarios based on the proactive processing and response of SDN switch requests by multiple controllers, relying on OpenFlow's Equal role. We also develop a proof-of-concept implementation in real and emulated switches to show the analysis of the recovery time from failures. Our results reveal that the proactive approach can reduce the average recovery time for controller failures by three orders of magnitude --- down to 10 milliseconds in the evaluated scenarios.