On Throughput Optimization in Optical Networks With Regular and Arbitrary Topologies

Abstract

We present some recent results regarding throug hput optimization in logically rearrangeable multihop lightwave networks. Two cases are distinguished: networks having arbitrary as well as regular connectivity topology. We use the term "arbitrary topology" in contrast to "regular topology", even though the degree of all nodes is fixed and the same in both cases. In both cases we present formulation of the combined station assignment/flow routing problem with the congestion minimization objective. We then outline a heuristic solution strategy based on tabu search. In terms of small congestion (as obtained heuristically), the results suggest that with increased problem sizes, regular topologies become more attractive. In such cases the benefit of having less restricted arbitrary network topology might not be fully utilized. Moreover, such design will implicitly offer benefits associated with management and rearrangement of regular topologies. We further experimented with a combined approach whereby initial assignments to network stations are restricted to a Perfect Shuffle connectivity pattern and obtained by solving a quadratic assignment problem. (In an earlier study, Perfect Shuffle proved to be the most promising among different regular topologies.) Following a transformation of variables, arbitrary connectivity patterns were allowed, and further improvements were performed. Computational results confirm the possible merit of this approach.