Traffic Engineering in Next Generation Networks Using Genetic Algorithms

NGN (next generation network) architectures are constantly evolving towards solutions that allow the operator to provide QoS-guaranteed services in heterogeneous, multi- domain and multi-services networks. In this context, one of the most advanced technologies is DiffServ-aware Traffic Engineering. It performs traffic engineering (TE) in a differentiated service environment by applying routing constrains with class granularity. This approach relies on the definition of an efficient bandwidth constraint model to allocate the available link resources in a per class type (CT) basis according to a certain set of resource utilization constraints. In this work, we describe a genetic algorithm (GA) that is aimed at finding the optimal bandwidth constraint settings for each link in the network. According to the TE objectives, the optimality is expressed in terms of a new cost function which weighs both the uniformity in network resource utilization and the average label switching path (LSP) length. The effectiveness of the proposed solution is analyzed comparing its bandwidth blocking rate and QoS results with those of two default bandwidth constraint configurations.