Intelligent congestion controller for high speed computer networks

Abstract

Quality of service Qos from a network depends mostly on the type of congestion controller available in the network Mostly congestion will occur due to mismatch in both arrival and departure rate to routers and the congestion will increase as network speed increases As a result the situation demands for effective congestion controller for today s high speed networks To address this situation in the present work fuzzy logic has been used to act as a congestion controller to operate computer networks at its best The fuzzy model has been developed based on queue drops current utilization sending rates of the senders bandwidth the aggregated incoming traffic and finally with waiting time This work introduces a novel concept to predict TCP protocol congestion which has not been dealt by researchers in earlier days The customers can realize the congestion only when they receive three or more than three acknowledgements or retransmission time out For prediction neural network has been used and in which training has been carried out using roundtrip time and TCP throughput as a neural network inputs Due to this a decision boundary is formed and it majorly predicts the congestion happening in computer networks This research also presents a controller design based on artificial neural network to avoid congestion in computer networks The quality of service for a network is not guaranteed by conventional controller due to the presence of non linearity and uncertain characteristics. newline