PERFORMANCE INVESTIGATIONS OF SOME ACTIVE QUEUE MANAGEMENT TECHNIQUES USING SIMULATION

Congestion is a fundamental problem in computer networks. Congestion may occur at router buffers of networks when the number of incoming packets cannot be absorbed by the available network resources such as buffer spaces. A poor network performance can be offered due to congestion, e.g. high dropping and queueing delay for packets, low throughput and unmaintained average queue length which may not prevent the router buffers from building up, then dropping packets. Different Active Queue Management (AQM) methods have been proposed to deal with congestion such as Random Early Detection (RED) and its variants like, Gentle Random Early Detection (GRED), REDD1 (the FLC method based on RED), etc. This paper presents a comparison between three AQM methods, i.e. GRED, REDD1 and the Adaptive GRED regarding to several performance measures such as: mean queue length (mql ), throughput (T ), average queueing delay ( D), overflow packet loss probability ( L P ) and packet dropping probability before the router buffer is overflowed ( p D ). This comparison evaluates the three AQM methods performance and then decides out which method offers more satisfactory performance measure results. The decision which method produces more satisfactory performance measure results is made solely depending on the varying of packet arrival probability parameter. This comparison shows that the Adaptive GRED offers the most satisfactory mql and D results when congestion occurs. Also, in presence of congestion; REDD1 loses the fewer packets due to buffer overflow and GRED drops the smallest number of packets before the router buffer has overflowed.