On Routing and Wavelength Assignment in Wdm Networks

With the help of advanced technology, WDM networks nowadays can support up to more than 100 wavelength on a single fiber. This feature brings new challenges to routing and wavelength assignment (RWA) problem, which is a NP-hard problem. There are three existing categories of integer linear programming (ILP) formulations to solve this problem optimally, namely, link-based, path-based and maximal independent set (MIS)-based formulations. All of the these experience bottlenecks of either wavelength symmetry problem or enormous problem size, which limits their scalability to very small networks. In this dissertation, we study the RWA problem in depth and greatly improve all three existing ILP formulations. For ring networks, we propose a new, exact ILP formulation that decomposes the path set recursively into several partitions and represents the original MIS with smaller MIS calculated in these partitions. As a result, this MIS decomposition formulation decreases the number of decision variables to a large extent, without sacrificing the benefit of MIS-based formulation. By our simulation analysis, compared with the existing formulations, the new MIS decomposition formulation gains a more-than-3-orders-of-magnitude decrease in running time. Typically, this new formulation based on MIS ILPs could solve any existing WDM ring networks with any number of wavelengths in just a few seconds. In mesh networks, we introduce the concept of a symmetric RWA solution and develop a new, fast path-based formulation for obtaining such solutions. Experimental study indicates that the approach improves the performance of the original path-based formulation by 2 orders of magnitude in terms of running time. With this new proposed formulation, we can solve mesh topologies representative of backbone and regional networks. The scalability of this approach makes it possible to characterize the performance of heuristics for the RWA problem and our experimental study shows that solutions obtained by the proposed approach are 20% − 60% better than those obtained by heuristics. As link-based formulation has a natural advantage over path-based and MIS-based formulation in terms of optimality in mesh networks, we propose a new, fast, link-based formulation with wise link selection algorithms. This new approach reduces the problem size by pruning the redundant link decision variables. The formulation scales well to mesh topologies representative of backbone and regional networks. Simulation results show that it decreases the running time by 2 orders of magnitude while providing better-quality solutions than path-based and MIS-based formulation. More importantly, this approach is applicable not only to the link-based …