Improved random early detection congestion control algorithm for internet routers

Abstract

In the internet, router plays a strategic role in the transmission of data packets. Active queue management (AQM) aimed at managing congestion by keeping a reduced average buffer occupancy and hence a minimal delay. The novel random early detection (RED) algorithm suffers from large average buffer occupancy and delay shortcomings. This problem is due in part to the existence of a distinctive linear packet drop function it deploys. In this paper, we present a new version of RED, called improved RED (IMRED). An important strategy of IM-RED is to deploy two dropping functions: i) nonlinear (i.e. quadratic) to deal with both light-and moderatenetwork traffic load conditions, and ii) linear to deal with heavy traffic load condition. Simulation experiments conducted using open-source ns-3 software to evaluate and compare the functionality of the proposed IM-RED with other two previous AQM algorithms confirmed that IM-RED reduces the average buffer occupancy and obtained an improved delay performance especially at heavy network traffic load scenario. Very fortunately, since RED algorithm is known to appear as a built-in model in ns-3 and even Linux kernel, its implementation can therefore be leveraged to obtain IMRED while only adjusting the packet dropping probability profile and holding on to its other attributes.

Developed the EX-RED Algorithm to Find Best Parameters Based on Load Connection Factor

“…An extension of DGRED is Stabilized Dynamic GRED "SDGRED" [13,14]. The Improved Random Early Detection I-RED algorithm [15] uses linear and nonlinear probability dropping packets. The I-RED modifies the default RED by dividing the threshold into two sections, which means three thresholds use minth, maxth, and between them to distinguish between two types of load traffic (light and heavy load).…”