HRED: a simple and efficient active queue management algorithm

Abstract

Active queue management (AQM) is an area of critical importance for the operation of networks. In this paper, we propose a minimal adjustment to the classic random early detection (RED) algorithm, called hyperbola RED (HRED), that uses the hyperbola as the drop probability curve. The control law of HRED can regulate the queue size close to the reference queue size which is settable by the user. As a result, it is expected that HRED is no longer sensitive to the level of network load, its behavior shows low dependence on the parameter settings, and it can achieve higher network utilization. Additionally, very little work needs to be done to migrate from RED to HRED on Internet routers because only the drop profile is adjusted. We implemented HRED on a real Internet router to examine and compare its performance with the classic RED and parabola RED that are currently deployed on Internet routers. From experiments on the real network, we conclude that HRED is insensitive to the network conditions and parameter settings, and can achieve higher network utilization than the other RED schemes