Distributed Software Defined Networking Controller Failure Mode and Availability Analysis
Paul Kevin Reeser, Guilhem Tesseyre, Marcus Callaway · 2019
Given the critical role Software Defined Networking controllers play in cloud computing and networking architectures, understanding their resiliency profile is crucial. Using OpenContrail as a reference architecture, we analyze the typical distributed controller failure modes and their effects on the control and data planes. We then develop hardware- and software-centric theoretical availability models for a variety of physical topologies and software modes of operation. These parametric models are used to predict availability and quantify sensitivity to underlying platform and process resiliency. The results suggest that the distributed control plane can achieve very high availability, while the host data plane may achieve much lower availability due to inherent single points of failure.