Offline multi-class flow allocation in MPLS networks using a distributed multi-objective genetic algorithm

The explosion of Internet-like data traffic and services (voice, video, multiplayer games, download, etc.) is fueling the demands for higher mobile telecom network capacities. Traditionally, mobile operators are used to add new transmission links and equipments to satisfy the traffic increase. However, this solution is firstly expensive in terms of capital and operation and secondly inadequate with the new telecommunication generations as 3G and 4G. In this paper, we propose a traffic engineering based approach to maximize the network usage. Based on an advanced Multi-Objective, Multi-Population Genetic Algorithm, we propose an offline distribution of Label Switch Paths (LSPs) that takes into consideration the scalability of the network, the Quality of Service (QoS) constraints for 4 Classes of Service (CoS) and recovery path to ensure high network availability. Congestion reduction, bandwidth optimization and other objectives have been combined to offer the network planner an optimal network management.