Towards improving resilience of controller placement with minimum backup capacity in software defined networks

The number of controllers required to manage the network and locations of these controllers affect various aspects of the separated controller plane including the performance metrics and ability to respond to failures. In this paper, we propose an optimization model for deploying controllers and mapping switches to controllers while achieving full resilience against a pre-specified number of controller failures. It minimizes the cost of the network while ensuring that each switch is mapped to a pre-specified number of controllers. Each switch sends only a fraction of demand to each of its assigned controllers. We also propose an optimization model to minimize the extra backup capacity reserved at controllers while fixing the total number of controllers in the network. Moreover, we introduce models to minimize the average and worst case switch to controller latency with resilience against controller failures. Our proposed models are evaluated using two widely used networks from Internet Topology Zoo. Evaluation results show that the proposed models necessitate a fewer number of controllers and achieve at least 50% reduction in the reserved backup capacity when compared to the existing approaches. Further, the proposed schemes achieve a higher resilience against controller failures and a lower coefficient of variation with respect to switch to controller latency when compared to the existing approaches. (C) 2018 Elsevier B.V. All rights reserved.