Joint scheduling and virtual topology design for sliding periodic traffic

Abstract

Two possible approaches can be considered for solving the virtual topology design problem for periodic (multi-hour) traffic demands. The first approach attempts to design a static topology that can accommodate all the traffic variations over time. The second option is to determine an appropriate series of virtual topologies to accommodate the different traffic loads at different times. This can lead to some savings in terms of the number of transceivers needed, but it requires the use of costly reconfigurable switching equipment. So, strategies for stable virtual topology design have received considerable attention in recent years. However, all the works reported in the literature so far, focus on the fixed window scheduled traffic model, where the start and end times of the demands are known in advance. In this paper, we propose a new integrated approach using the more general sliding window model, for jointly scheduling the demands in time and designing a logical topology that can accommodate all the scheduled demands. The goal is to a find a suitable static topology that can handle fluctuations in the offered sub-wavelength traffic load, without requiring the use of reconfigurable optical switching equipment. We first present a comprehensive integer linear program (ILP) formulation for designing a cost-efficient, stable logical topology for time-varying demands, and then propose an integrated heuristic algorithm capable of handling larger networks. Simulation results demonstrate the advantages of the proposed approaches, not only compared to holding time unaware models, but also over the traditional fixed window model.