Avaliação de diferentes configurações do algoritmo RED para a diminuição da auto-similaridade do tráfego de rede

Abstract

With the knowledge that the network traffic presents the characteristic of self-similarity, many studies have been carried out seeking its decrease, since the self-similarity in the traffic causes negative effects, like increase in the buffer rate, delay in the queues and network congestion. One of the factors most addressed by researchers is queue management policies, including RED and droptail, due to the ease of configuring the algorithms in the routers. In this work, with the help of the NS3 and R tools, we developed a standard to help in the configuration of RED and show the impact of different configurations of the RED in the network, being possible to observe that some configurations are better, depending on the load of traffic of the network, having the self-similarity and packet loss rate lower than the others. In addition, we perform RED comparisons with the droptail showing that the performance of both is quite similar.