Active queue management for queue stability and fairness

Abstract

TCP flows tend to iterate increasing their traffic to congest link and decreasing to respond congestion. The characteristics of this predictable behavior can be exploited to detect traffic structure and enhance the system performance. We propose a new active queue (AQM) scheme to reduce queue oscillation and realize stable queue length. The proposed scheme measures the current arrival and drop rates, and uses them to estimate the next arrival rate. Based on this estimation, the scheme calculates the drop probability which is expected to realize stable queue length. Secondly, the proposed AQM is expanded to a fairness-enforcing AQM scheme controlling both the fairness and link congestion harmonically. The expansion is constructed by embedding fairness-enforcing algorithm into the proposed AQM scheme. The scheme provides almost max-min fair bandwidth to heterogeneous flows and also keeps very stable queue size. The complexity is very low by avoiding redundant state of small flows with a novel high-bandwidth flow detection algorithm. The evaluation results compared to the other well known schemes, demonstrates the effectiveness of the proposed scheme. We also propose a minimal overhead AQM scheme which approximates max-min fairness without keeping any per-flow state. The designed algorithm is very simple but effective to enforce max-min fair share among flows. The proposed scheme uses very light resources including additional small memory. The evaluation result verifies the max-min convergence of the proposed mechanism and shows that the proposed scheme outperforms the compared related schemes.