Applying computational intelligence for congestion avoidance of high-speed networks

Congestion avoidance is one of the key issues for the effectiveness and reliability of high-speed networks. The design of congestion avoidance mechanisms is usually based on some assumptions of the traffic model and the performance is strongly influenced by the related parameters. Such parameters are usually initialized subjectively and adjusted by applying some predefined linear transformation to the current value, which is not optimal and often leads to poor performance. In this paper we propose a design principle for congestion avoidance in high-speed networks, which utilizes computational intelligence such as fuzzy logic and genetic algorithm as design tools. Our GAFuCA model has the ability to design optimal congestion avoidance schemes without the precise knowledge and mathematical model of data traffic. Furthermore such schemes can be optimized according to different criteria. GAFuCA model has been used to design a genetic-fuzzy-logic-based RED algorithm in TCP/IP networks, FuRED. Simulation results have shown that a better performance can be reached by applying such a design process.