Multi-objective traffic grooming in WDM network using NSGA-II approach

This paper considers a multi-objective network design problem for the traffic grooming, routing and wavelength assignment (GRWA) in WDM networks. The design objectives are to maximize the number of accepted communication requests (source-destination pairs) as well as to minimize the number of wavelength channel requirement. Both the design objectives are conflicted to each other; maximizing the number of accepted commodities will require a large number of wavelength channels while minimizing the number of wavelength channels will limit the amount of accepted commodities. To solve the multi-objective network design problem, we apply a fast and efficient optimization technique called "Fast Non-dominated Sorting Genetic Algorithm (NSGA-II)". In this paper, traffic grooming (GA-LMF) and non-traffic grooming (GA-MDF and FAR-FF) algorithms are compared and benchmarked for solving the multi-objective design problem. The results show that the GA-LMF is the most flexible and efficient grooming technique. The obtained solutions from the GA-LMF are spread on the objective space and are better than those from other non-grooming techniques in both objective values (i.e., number of accepted commodities and wavelength channels required).