pathWave: Combined estimation of network link capacity and available bandwidth using statistical signal processing

Network link capacity (lc) and available bandwidth (av-bw) are recently considered as importance for various network domains. Consequently, several lc and av-bw measurement techniques and tools have been proposed, where empirical experience, performance comparison and foundational understanding of lc/av-bw measurement have been focused. Many heuristic or non-probabilistic statistical algorithms have been used to estimate lc/av-bw in this literature. However, stochastic analysis indicates that improved understanding of the fundaments of bandwidth estimation could lead to more accurate and effective methods. In this paper, we model network lc/av-bw measurement as a stochastic signal state detection and signal parameter estimation problem. We present a novel approach to measure network lc/av-bw, using packet-pair dispersion technique and our extended self-induced congestion principle based on statistical signal processing theory. We validate the affectivity of our lc/av-bw measurement tool, pathWave in NS-2 simulation environment and evaluate its performance in test bed by comparing with other representative tools. According to the analytical, simulative, and experimental result, we can demonstrate that our statistical signal processing is a more efficient approach for lc/av-bw measurement.