SRF: Random Expanders for Designing Scalable Robust and Fast Communication Networks

Abstract

A generalized approach to obtain random expander graphs is proposed, which includes the growth of the network by adding to it any suitable random regular network, iteratively, in special cases, a node, an edge. We show that the proposed algorithm can produce good expanders (Ramanujan graphs) that are used to design fast, scalable communication networks. The qualitative and numerical analysis of the produced communication networks is performed on the basis of predefined structural and spectral measures and metrics, for example, mean-first-passage-time, eigenratio, clustering coefficient, and average path length. Apart from that, a simple message passing communication protocol is simulated over the proposed growing expander graphs and other state-of-the-art network topologies, BAM, ERM, SWN, CSWN and SWRN models, and delay is calculated. Results show that the constructed random expanders have low network latency, high convergence rate and robustness. IEEE