Link Loss Rate Inference Using Success Rate Cumulant Generating Function

Inference of the internal link state is an important and challenging issue for operating and evaluating networks. This paper presents a method to infer internal link loss characteristics based on end-to-end measurement. Our method uses cumulant generating function (CGF) inference algorithm. The main contribution of our approach is that we use the success rate CGF instead of the loss rate CGF, because the loss rate CGF cannot be constructed directly. We construct the path success rate CGF first, then the link success rate CGF can be inferred, and the link success rate can be obtained. Employing the relationship between the link loss rate and the link success rate, we can get the link loss rate. The simulation results demonstrate that this method is efficient.