Congestion Control in Diffi-serv Using Fuzzy and Neural Methodology

The paper covers a brief introduction of the problem of congestion and its control using the rule base of the fuzzy module refined by neural module. A number of new applications being developed for the Internet require stringent bounds on parameters such as the end-to-end delays and inter-packet jitter. In order to provide these bounds led to the development of alternate mechanisms based on policies that could be deployed on the existing network infrastructure with minimal changes, and still provide a suitable Quality of Service. Differentiated Services (Diff-Serv) is one such technology. DS can be implemented with relatively small disturbance to the existing infrastructure. For this the congestion control algorithm called Random Early Detection is discussed with fuzzy-neuro system as the technique called Random Early Detection technique used in differentiated services network by applying Neuro-Fuzzy approach is slightly modified. First, Introduction to fuzzy inference module by applying various fuzzy based control strategies is there and then refined with its rule-base with the help of neural network to incorporate network dynamic conditions which keeps on changing with time. The experience with NS2 network simulator is also discussed.