Performance of Active Queue Management Over Multiple Congested Links

Recent studies have shown that a non-negligible number of packets face multiple congested links on Internet paths. We investigate the impact of multiple congested links on Active Queue Management (AQM) techniques. We present the results of an ns-2 evaluation study of various AQM techniques that, in contrast to previous studies of AQM, uses a complex network topology including up to 3 congested links, reverse path traffic, and realistic round-trip times. We look at the effects of multiple congested links on the throughput of long-lived FTP flows and HTTP response times of web traffic. We found that overall, performance degrades as congested links are added, but our main conclusions hold regardless of the number of congested links. In our experiments we found that drop-tail queuing provides the best performance for long-lived flows, but is the worst performer for HTTP traffic. Adaptive Random Early Detection (ARED) provides the worst FTP performance, with short queues at the cost of high packet loss. Both the Proportional Integral controller (PI) and Random Exponential Marking (REM) provide reasonable FTP throughput, low HTTP response times, and very low packet loss.