Robust Identification of Shared Losses Using End-to-End Unicast Probes
Robust Identification of Shared Losses Using End-to-End Unicast Probes
Multicast-based inference of network-internal loss characteristics
IEEE Transactions on Information Theory
Using Bayesian network on network tomography
Computer Communications
| Hi-index | 0.00 |
Network tomography aims to obtain link-level characteristics, such as loss rate and average delay of a link, by end-to-end measurement. A number of methods have been proposed to estimate the loss rate of a link by end-to-end measurement, all of them, in principle, are based on parametric statistics to identify unknown parameters. In addition, they all used the traditional centralized data processing techniques to complete the estimation, which is time-consuming and unscaleable. In this paper, we put forward a distributed method to tackle the scalability problem. The proposed method, instead of estimating link-level characteristics directly, estimate path level characteristics first that can be executed in parallel and can be achieved by a distributed system. The path level characteristics obtained can be used to identify link-level ones later. The proposed method has been proved to be an maximum likelihood estimate.