Ground Control to Major Faults: Towards a Fault Tolerant and Adaptive SDN Control Network

To provide high availability and fault-tolerance, SDN control planes should be distributed. However, distributed control planes are challenging to design and bootstrap, especially if to be done in-band, without dedicated control network, and without relying on legacy protocols. This paper promotes a distributed systems approach to build and maintain connectivity between a distributed control plane and the data plane. In particular, we make the case for a self-stabilizing distributed control plane, where from any initial configuration, controllers self-organize, and quickly establish a communication channel among themselves. Given the resulting managed control plane, arbitrary network services can be implemented on top. This paper presents a model for the design of such self-stabilizing control planes, and identifies fundamental challenges. Subsequently, we present techniques which can be used to solve these challenges, and implement a plug & play distributed control plane which supports automatic topology discovery and management, as well as flexible controller membership: controllers can be added and removed dynamically. Interestingly, we argue that our approach can readily be implemented in today's OpenFlow protocol. Moreover, our approach comes with interesting security features.