WDM Optical Networks: A Complete Design

Abstract

Different integer linear programming (ILP) have been proposed for design of optical networks. The traditional approaches divide design into two separate problems: virtual topology design (VTD), in which best connections among nodes are derived from traffic demand; and routing-and-wavelength assignment (RWA), in which physical paths are accommodated in the physical topology to support the requested connections. We propose an iterative linear programming approach to solve both problems jointly under multiple objectives such as congestion avoidance, fiber load and wavelength pool minimization. The solution of the VTD problem generates a request for a set of paths to be supplied by the physical topology. Physical paths are then allocated in order to minimize some objective functions that are akin to a linear programming formulation. If no feasible solution is found, VTD program supplies a next best solution until all paths are routed. Some objective functions (e.g. maximum fiber load) may be oblivious to the persistence of cycles in the final solution, which may even be dismembered from the source-to-destination link sequence. These anomalies may be eliminated by re-optimizing the solution using the total number of hops as a new objective function, subject to the minimal value of maximum fiber load that was determined in the previous optimization step. The final design phase is the assignment of wavelengths to paths or sections thereof, making best use of available wavelength conversion resources. Our formulation allows for any kind (partial or full, sparse or ubiquitous) of wavelength conversion and limited number of converters, thus providing a tool for the allocation of conversion resources in the network.