Load-balancing on swapped or OTIS networks

Existing local iterative algorithms for load-balancing are ill-suited to many large-scale interconnection networks. The main reasons are complicated Laplace spectrum computations and flow scheduling strategies. Many large-scale networks are modular and/or hierarchically structured, a prime example being the class of swapped or OTIS networks that have received much attention in recent years. We propose a new local scheme, called DED-X, for load-balancing on homogeneous and heterogeneous swapped/OTIS networks. Our scheme needs spectral information only for the much smaller basis or factor graph, which is of size O(n)O(n) rather than O(n2)O(n2), and it schedules load flow on intragroup and intergroup links separately. We justify the improvements offered by DED-X schemes over traditional X schemes analytically and verify the advantages of our approach, in terms of efficiency and stability, via simulation.