A Hybrid SDN Path Computation for Scaling Data Centers Networks

Initial SDN deployments endorsed a centralized architecture with a single controller that is responsible for managing entire networks. Yet, this scheme has proven to be unscalable and unreliable. Though distributed control plane architectures have gained ground recently, they bring new concerns. A principal question that should be investigated thoroughly is how to efficiently perform path computation and update flow tables in very large SDNs. In this paper, we address this problem by proposing ScalCon, a scalable hybrid SDN architecture that distributes the control plane over different controller types (i.e., Fog, Edge, Cloud). ScalCon distributes the load of path computation among all controllers and uses a multi-priority queening model and node parallelism to cut down the synchronization overhead. Simulation results show that ScalCon outperforms existing schemes in terms of path computation time, path setup latency. end-to-end delay and communication overhead.