Towards an Adaptive RED Algorithm for Achieving Delay-Loss Performance

Random Early Detection (RED) [1] is an active queue management strategy that has been deployed extensively in routers. However, its effectiveness has been limited owing to the difficulty in tuning its parameters under different network conditions. In this paper, we define the notion of an operating point that an Active Queue Management (AQM) policy should target. We suggest that the RED parameters should be set or varied with the aim of reaching this operating point. We have performed extensive simulations to understand the effect of maxth and maxp on the network dynamics. Based on these simulations, we give guidelines to adapt the RED parameters.