Proactive controller assignment schemes in SDN for fast recovery

Abstract

Software-defined networking (SDN) is an emerging networking paradigm which entails separate control and data planes. In SDN, a controller is a centralized network control entity responsible for management of the data plane. Typically, a sizeable software defined network with a single controller responsible for all forwarding ele ments is potentially failure-prone and inadequate for dynamic network loads. To this end, having multiple controllers improves resilience and distributes network control load. However, when there is a disruption in the control plane, a rapid and perfor mant controller-switch assignment is critical, which is a challenging technical question. Therefore, controller-to-switch assignment and robust operation are challenging re search questions, since control plane hazards such as disconnection between control and forwarding planes may lead to packet loss and failures in the system. In this the sis, we propose a proactive switch assignment scheme PREF-CP in case of controller failures using genetic algorithm considering controller load distribution, reassignment cost and probability of failure. Moreover, we compare the performance of our scheme with other algorithms developed during our thesis work, namely random and greedy algorithms. Experiment results show that our proposed PREF-CP framework has better assignment performance in terms of probability of failure and controller load distribution.