A Differential Evolution with Pareto Tournaments for solving the Routing and Wavelength Assignment problem in WDM networks

Abstract

The technology based on Wavelength Division Multiplexing (WDM) applied to optical networks has resolved the bandwidth waste in this kind of networks. WDM divides the bandwidth of an optical fiber in different wavelengths that can be used by electronic devices to send and receive data without bottlenecks. Another problem appears when the necessity of choice of the path and the wavelengths to interconnect a set of source-destination pairs comes up. This problem is known as Routing and Wavelength Assignment (RWA) and there are two types, depending on the demands: Static-RWA and Dynamic-RWA. In this paper we present a multiobjective evolutionary algorithm to solve this problem. We choose the Differential Evolution (DE), incorporating the concept of Pareto Tournament (DEPT). To determine the parameters of the algorithm, we used two real different topologies (the first is a topology from USA, NSF network; and the second is a topology from Japan, NTT network) and six sets of source-destination pairs for each topology, that is, a total of twelve instances. After all experiments, we can conclude that with this multiobjective evolutionary algorithm, we have obtained better results than the other approaches published in the literature.

GRASP for traffic grooming and routing with simple path constraints in WDM mesh networks

“…Our GRASP algorithm 522 and LocalSolver are run for 10 times for each instance. G G20_200_1 20 37 200 32 G20_200_2 20 39 200 32 G20_200_3 20 24 200 32 G20_200_4 20 39 200 32 G20_200_5 20 35 200 32 G40_200_1 40 80 200 32 G40_200_2 40 105 200 32 G40_400 40 90 400 32 G100_500 100 300 500 32 We run three versions of our algorithms for this set of 555 instances and the time limit is set to be 2 h. The detailed 556 computational results are shown in Table 6 Two optical topologies used in the literature [14,31] are 571 tested in this section. These two networks are called the 572 European optical network (COST239) and the National Sci-573 ence Foundation network (NSF).…”

“…This is referred to as the routing and 13 wavelength assignment (RWA) problem [1]. In order to maxi- 14 mize the usage of the lightpaths, telecommunication carriers adopt a technique that consists of efficiently grooming low 16 speed traffic streams into high capacity channels. This tech-17 nique is referred to as the RWA problem with traffic groom-18 ing (GRWA).…”

Discrete multi-objective differential evolution algorithm for routing in wireless mesh network

“…Routing and Wavelength assignment problem using Differential Evolution with Pareto Tournaments (DEPT) was presented by Rubio-Largo et al (2010). The objectives are hop count and number of wavelength conversion.…”

The software project scheduling problem: A scalability analysis of multi-objective metaheuristics

“…The fundamental idea behind DE is a scheme for generating new possible solutions (trial individuals) taking advantage of the differences among the population (target individuals), according to its simple formulae of vector-crossover and mutation. We have defined a new multiobjective version that incorporates the Pareto Tournaments concept (DEPT) [16] to choose the best solution between two given ones based (in this case, the target and the trial individuals). Ties are broken in the tournament (in case on nondominance) by using the crowding distance of NSGA-II.…”

“…But actual, relevant, contemporary software projects involve both more people and more tasks. This work evaluates the issue of the scalability of eight multi-objective metaheuristics for the SPS problem, three classical methods -NSGA-II [13], SPEA2 [14], and PAES [15]-plus five novel algorithms -DEPT [16], MO-FA [17], MOABC [18], MOCell [19], and GDE3 [20]-on a set of 36 instances with an exponential increase in both the number of tasks (from 16 to 512) and the number of employees (from 8 to 256). Two quality indicators, the hypervolume (HV) [21] and the attainment surfaces [22], have been used to measure the quality of the resulting Pareto fronts.…”