An adaptive prediction based approach for congestion estimation in active queue management (APACE)

Active queue management (AQM) policies are mechanisms for congestion avoidance, which pro-actively drop packets in order to provide an early congestion notification to the sources. We propose a new adaptive prediction based approach for congestion estimation (APACE) in active queue management that predicts the instantaneous queue length at a future time using adaptive filtering techniques. We compare the performance of APACE with other existing AQM schemes like RED (random early detection), SRED (stabilized RED), AVQ (adaptive virtual queue) and PAQM (predictive AQM) in networks having both single and multiple bottleneck links. We show that the performance of APACE in terms of the stability of the instantaneous queue is comparable to that of PAQM, which is one of the best AQM strategies. Moreover, APACE performs better than PAQM in terms of link utilization and in networks with multiple bottleneck links. APACE is not very sensitive to parameter settings, adapts quickly to changes in traffic and can achieve a given delay or packet loss by varying just one parameter, thus giving the network operator a better tool to manage congestion and improve the performance of the network.