AmRED and RED-QE: redesigning random early detection algorithm

Over the last few decades ago, the study of active queue management (AQM) has resulted in numerous improved algorithms for performing congestion control at the router. The most influential AQM algorithm, however, remains Random Early Detection (RED). A significant problem with RED is that the drop policy component implemented therein is characterized by the use of a solitary linear function, which alone is insufficient for combating network congestion as it ends up in large delay when heavy traffic workload occurs. To overcome this problem, therefore, in this paper, the packet dropping policy of RED is revisited with a focus on designing two new alternative approaches, namely, Amended RED (AmRED) and RED-Quadratic Exponential (RED-QE). Simulation evaluations were conducted in a discrete-event simulation platform (i.e., ns-3) using different scenarios to prove the effectiveness of AmRED and RED-QE against two contemporary AQM algorithms. Simulation reports suggested that RED-QE exhibits superiority. Both AmRED and RED-QE algorithms are amenable to implementation in network routers, considering that only the packet dropping probability profile of RED’s implementation requires very little rectification.