An Optimized Type-2 Fuzzy Logic Model Based on Weighted Random Early Detection for Congestion Detection in Wired Networks

Abstract

Congestion is the greatest challenge facing the transmission of data packets across computer networks. Congestion in the router buffer increases packet loss. Packet loss is a major problem of networks and their control and reduction are objectives for various management techniques. A congestion control management mechanism is operated on the router to respond to congestion when it occurs. Active Queue Management (AQM) methods are able to detect congestion at an early stage and control it by packets dropping. The Weighted Random Early Detection (WRED) method, amongmany other AQM methods, gives good performance to detect, and control congestion, as well as preserve packet loss. In this thesis was presented a methodology to building a model to simulate how the WRED algorithm works, and improve it by combining it with Type-2 fuzzy logic system, where experiments are simulated under the discrete time queue in Java and OMNeT++ IDE to collect a data-set containing the queue sizes, average queue lengths and the number of packets dropped. The Type-2 Fuzzy Logic System was presented a model that is very suitable for the handling of the encountered uncertainties in the input factors, and also converting the accumulated data to linguistic formats which can be easily stored and analyzed. In this thesis, the parameters of fuzzy membership functions of Type-1 and Type-2 Fuzzy Logic Systems are extracted dynamically from data-sets using the Fuzzy C-Means (FCM) clustering technique and they learn the rules from data. The results show that the proposed type-2 Fuzzy Logic-based Systems (FLS) outperformed type-1 FLS with their WRED counterparts, according to dropping and packet loss rates. The packet loss has been reduced at the rate of 32%compared to WRED, depending on the arrival probability. The Big Bang-Big Crunch (BBBC) algorithms was employed to optimize the proposed Type-2 Fuzzy Logic System (FLS) parameters for fuzzy logic membership functions and the number of fuzzy rules, to allow best execution to detect and control congestion. The results show that the BB-BC Type-2 FLS outperformed Type-2 FLS and Type-1 FLS with their WRED counterparts, according to dropping and packet loss rates. The packet loss has been reduced at the rate of 35% compared to WRED