Empowering self-diagnosis with self-modeling

This paper proposes an approach to automatise the management of faults, covering the different segments of a network, and the end-to-end services deployed over them. This is model-based approach addressing the two weaknesses of model-based diagnosis namely deriving an accurate model and dealing with huge models. To address the first point, we propose a solution called self-modeling that formulates off-line generic patterns of the model, and identifies on-line the instances of these patterns that are deployed in the managed network. The second point is addressed by an active (self-)diagnosis engine, based on a Bayesian network formalism. This consists in reasoning on a progressively growing fragment of the network model: more observations are collected and new tests are performed until the faults are localized with sufficient confidence. This active diagnosis approach is experimented to perform cross-layer and cross-segment alarm management on an IMS network.