Analysis of selection and crossover methods used by genetic algorithm-based heuristic to solve the LSP allocation problem in MPLS Networks under capacity constraints

Abstract

The Multiprotocol Label Switching (MPLS) is a popular routing technique for IP networks, where the core problem is to find a route (called LSP) that satisfy all the capacity constraints imposed by a specific traffic. Genetic algorithms come as a simple, appealing solution approach, but one that requires careful choices concerning initial population generation, crossover, mutation and selection. The present paper discusses the influence of different crossover and selection methods in achieving a fast and accurate convergence of the genetic algorithm, when solving the MPLS allocation problem. The experimental results, using different network topologies such as Carrier, Dora, and Mesh, have shown that uniform crossover and Stochastic Remainder Sampling selection are the most suitable combination to solve the problem.