Statistical Detection of Congestion in Routers

Detection of congestion plays a key role in numerous networking protocols, including those driving active queue management (AQM) methods used in congestion control in Internet routers. This paper exploits the rich theory of statistical detection theory to develop simple detection mechanisms that can further enhance current AQM methods. The detection of congestion is performed using a maximum-likelihood ratio test (MLRT), which reveals that the likelihood of congestion grows exponentially with the queue occupancy level. Performance evaluation of the likelihood detector shows it is robust to variations of the network parameters. The mathematical expression of the likelihood of congestion depends on the router's current dropping rate, its desired queue occupancy level, and the current queue occupancy. When incorporated into random early marking (REM) and random early detection (RED), the likelihood-ratio-based detection considerably improves their reaction time and reduces the variance of queue occupancy values.