Placing Controllers over Complex Wide Area SDNs Based on Clique Identification

Abstract

Wide Area Networks (WANs) form the network core that covers wide geographical areas. WANs often have complex topologies, and it is challenging to incorporate multiple controllers in the control plane to reduce the network delay in Wide Area Software Defined Networks (WASDNs). We propose a distributed controller placement problem (DCPP) for various control plane structures to address this challenge. While existing exhaustive and greedy algorithms cannot efficiently solve the DCPP over many large-scaled WASDNs, we propose a network simplification strategy based on a novel global network coefficient, polyindex, to identify all the nonoverlapped cliques in networks and characterize the topology features of such complex networks. With such strategy, the good number, organization, and placements of controllers for the DCPP over large-scaled WASDNs can be determined. Extensive evaluations demonstrate the effectiveness of the polyindex in capturing the features of sparse WANs. While applying the proposed strategy over large-scaled WANs with small and medium polyindexes can quickly find the placements for the DCPP while meeting the given delay requirement, carefully adjusting the delay requirement and threshold is the key to generate high quality frontiers while keeping the time cost low over the WANs with large scales and polyindexes.