SDN Controller Robustness and Distribution Framework

Abstract

SDN improves network flexibility which is constrained by network protocol in a conventional network by decoupling the control plane and the data plane of the network. This is the reason why many companies and universities migrate their network to SDN, there will be more SDN network in the future. Yet SDN network mainly depends on the controller in the control plane. Hence, SDN controller robustness becomes an important issue, because a controller failure will result to a network outage.

OpenFlow is arguably the standard protocol for SDN network. Thus, it is necessary to investigate the robustness of the OpenFlow control plane. Several open source controllers such as OpenMul, Floodlight, Opendaylight, and ONOS have multiple controllers framework to tackle a controller failure. They provide failover mechanism, when there is a controller failure, a backup controller can take over to control the network. In this thesis a benchmark is conducted to measure how long the failover time of those open source controllers. Unfortunately their failover time is in order of seconds, which is way higher than 50ms, the acceptable standard of carrier-grade recovery time.

This thesis presents a solution that can improve SDN robustness: A Controller Robustness and Distribution Framework (ACRoDiF). ACRoDiF is compatible with several open source OpenFlow controllers such as Ryu, OpenMul, Floodlight, Opendaylight, and ONOS. ACRoDiF provides failover mechanism that has lower failover time than in the open source controllers: 76ms. It can also eliminate failover time completely if using two active primary controllers.