Metaheuristic Methods for Solving the Capacity and Flow Assignment Problem in TCP/IP Networks

This paper proposes the use of two metaheuristic methods, the GRASP (greedy randomized adaptive search procedure) and the PSO (particle swarm optimization), customized to TCP/IP networks dimensioning. Specially, this article treats the capacity and flow assignment (CFA) problem considering the packet delay allowed within every source/destination pair. Each router is modeled using a Markovian queue with batch arrivals, in order to consider the burst traffic yielded by TCP protocol. This study suggests still an iterative procedure to solve the network buffer dimensioning problem. Performance results achieved for some network topologies are faced with simulated results (achieved from NS-2) validating this proposed methodology.