Influence of Active Queue Management Parameters on Aggregate Traffic Performance
Abstract
Active queue management (AQM) is a family of packet dropping mechanisms for FIFO queues that have been proposed to support end-to-end congestion control mechanisms in the Internet. In this paper, we examine the four possible combinations of queue size averaging vs instantaneous queue size and a sharp dropping function vs a smooth dropping function. The combinations include both the well known "Drop from Tail" (TD) and "RED" mechanisms as well as the more recently proposed "Gentle RED" and a previously unrecognized mechanism we call "Gentle RED with instantaneous queue size", or GRED-I. The results show the end-to-end performance of the four AQM mechanisms. Thus, we do not study individual TCP flow performance but instead the flow aggregation behavior through backbone nodes, and we chose experimental parameters and observed metrics accordingly. We show that if TD and RED have similar behaviors in terms of aggregate TCP throughput, RED discards more non responsive traffic than TD. On the other hand, we show that RED creates more consecutive drops than TD. We identify queue size averaging and the slope of the dropping function as the two major parameters influencing consecutive drops.
Domains
Other [cs.OH]
Loading...