End-to-end inference of link level queueing delay distribution and variance

The statistical characteristics of the delay over each link of a network path are a worthy information for different purposes: troubleshooting, traffic engineering, adaptive multimedia flow coding, overlay network design, etc.. However, querying each node on the path in order to retrieve this kind of information can be unfeasible or just too resource demanding. For these reasons many algorithms have been devised in order to infer the internal state of a network based on end-to-end measurements: in fact, the inference of the delay distribution and variance is the focus of this paper. The algorithms which have been proposed in the literature rely on active measurements and are based on a single-sender multiple-receivers scheme, thus requiring the cooperation of a possibly wide number of nodes; notwithstanding, they do not guarantee the possibility of characterizing the delay on each hop of an end-to-end path. On the contrary, the techniques that we propose in this paper are intended to infer the delay distribution and variance over each link of a given network path, based on two-points measurements only. We assess the correctness of our algorithms with reference to widely accepted theoretical results and evaluate their performance through a wide series of model-based and ns2 based simulations.