RED-LE: A Revised Algorithm for Active Queue Management

Abstract

The random early detection (RED) algorithm was developed in 1993. Nearly three decades later, several improved variants have been proposed by scientists. The use of a (pure) linear function for computing packet drop probability has turned out to be a disadvantage, leading to the problem of large delays. Such a problem may be addressed by using linear and non-linear (i.e. as exponential) packet drop probability functions. This paper proposes a revised RED active queue management algorithm named RED-linear exponential (RED-LE). This variant involves an interplay of linear and exponential drop functions, in order to improve the performance of the original RED algorithm. More importantly, at low and moderate network traffic loads, the RED-LE algorithm employs the linear drop action. However, for high traffic loads, RED-LE employs the exponential function for computing the packet drop probability rate. Experimental results have shown that RED-LE effectively controls congestion and offers an improved network performance under different traffic loads.