STUDY AND IMPROVE NETWORK PERFORMANCE USING RANDOM EARLY DETECTION (RED) BY COMBINING FINE-TUNING THE LOWER THRESHOLD AND AVERAGE QUEUE SIZE

Abstract

Internet is expected to better support many applications such as multimedia applications with limited bandwidth, low delay and packet loss rate requirements. Therefore, there is a need for new mechanisms to control congestion in the network. Active queue management (AQM) algorithms play an important role in ensuring network stability. RED (Random Early Detection) is the first dynamic queue management technique implemented in TCP/IP networks for congestion avoidance control. RED is based on comparing the average queue length with lower and upper thresholds to mark or discard packets. RED strategies have been introduced with the adjustment of parameters of threshold, average queue size to obtain better network performance. In this paper, the author proposes an improved RED algorithm called TqRED (Threshold queue RED) to address the limitation of RED. Through the evaluation simulation on the NS2 simulator, the author found that TqRED gives better results than RED in terms of packet loss, average queue delay, and average throughput.