Local area network analysis using end-to-end delay tomography

There has been considerable interest over the last few years in collecting and analyzing internet traffic data in order to estimate quality of service parameters such as packet loss rates and delay distributions. In this paper, we focus on fast and efficient estimation methods for network link delay distributions based on end-to-end measurements obtained by probing the underlying. We introduce a rigorous statistical framework for designing the necessary probing experiments and examine the properties of the proposed estimators. The proposed framework and the resulting methodology are validated using data collected on the University of North Carolina (UNC) local area network.