Novel active queue management algorithm designed for self-similar traffic.

Self-similarity of network traffic has a great impact on network performance, and can lead to large queuing delays and packet loss rates. To avoid this problem, this paper proposed an active queue management algorithm considering the selfsimilarity of traffic. The proposed algorithm estimated based on the wavelet method, and simultaneously analyzed the self-similar process and determined the degree of the self-similarity using the estimation results. Then it classified the network traffic based on the real-time estimation of Hurst parameter. In the simulation, it generated the network configuration by MATLAB, and simulated the network traffic by Pareto distribution. The performance of the proposed algorithm was compared with that of other buffer management algorithm, and the simulation results verify that the proposed algorithm has better fairness, and show that under congested circumstances, the proposed algorithm can avoid the influence of dropped packets on other queues. [ABSTRACT FROM AUTHOR]