Most work on Differentiated Services (DiffServ) handles Quality of Service (QoS) provisioning on a per node basis, which assumes that this strategy would provide QoS in the whole domain. Nevertheless, this approach could fail in large domains with multiple flows aggregation and unexpected input traffic. Therefore, provisioning techniques should be used to avoid unpredicted overloads that result in QoS fluctuations. A proposal using fuzzy controllers to reconfigure DiffServ nodes according to

more »

... ress traffic and achieved QoS was presented in [1]. However, it is not easy to specify fuzzy rule bases and membership functions that optimize the controllers performance. Thus, we propose a methodology to choose optimized fuzzy controller parameters using the Wang-Mendel and genetic algorithms. Finally, we evaluate the performance of this methodology by simulation of voice over IP applications in DiffServ domains.