Active Queue Management for Self-Similar Network Traffic

Abstract: Recent studies have shown that network traffic is Self-Similar, and it has a great impact on network performance. Self-similar traffic can lead to large queuing delays and packet loss rates. In this paper, we devise an active queue management algorithm which takes the Self-similarity of traffic into account. Hurst is a key parameter describing self-similar processes, which is designed to determine the degree of the self-similarity. In our approach, we utilize a technique based on the wavelet method to estimate the Hurst parameter. Classification is based on real-time estimation of Hurst parameter. Also, we use ns2 to simulate the network configurations and to generate traffics with Pareto distribution. The numerical results illustrate the performance of the proposed algorithm in contrast to other recently implemented buffer management algorithms in ns2