A new model for optimal routing and wavelength assignment with fixed-length tunnel allocation in multi granularity cross-connect WDM networks

Abstract

Multi-granularity Optical Cross-Connect (MG-OXC) has been proposed to provide a cost-efficient way to support the growing demand for bandwidth. In the MG-OXC networks, consecutive wavelengths are bundled to form a tunnel and then switched as a single unit. Network resources at the intermediate nodes on the route of a tunnel, including switching fabrics and multiplexers, can thus be reduced. In this paper we consider the problem of routing and wavelength assignment (RWA) with tunnel allocation in MG-OXC networks. Given a set of static lightpath requests, our problem is to (a) allocate a set of fixed-length tunnels (b) find routes from the source nodes to their respective destination nodes, and (c) assign wavelengths to theses route. The objective here is to minimize the blocking probability. Furthermore, we extend our work for the dynamic RWA problem. Given a historical traffic matrix, the dynamic RWA problem is how to build a set of tunnels offline to accommodate the future dynamic lightpath requests in such a way that the blocking probability can be minimized. In order to utilize the wavelength ports and fibers efficiently, each tunnel established should follow a tunnel length constraint which could be equal to the average network hop distance. Based on this criterion, a novel graph model is proposed [7] in which edges are added only for the node pairs whose hop distance follow the tunnel length constraint to form an auxiliary graph. In this paper, we suggest a 0/1 Integer Linear Programming (ILP) formulaion for RWA with tunnel allocation in MG-OXC networks under the tunnel length constraint. The previous auxiliary graph model is extended to a layered auxiliary graph model to facilitate the formulation. We compare the performance of different heuristics, including CB-STA [1], WTA and PCWTA [7], to the ILP solution. The simulation results show that PC-WTA and WTA outperforms CB-STA in all switching type combinations and network topologies, and they are very close to the optimal value calculated from our ILP formulation.