Network Tomography with Partial Topology Knowledge and Dynamic Routing

Abstract Networks are always evolving to meet the needs of progressing and novel applications. To this end, improved network capacity, latency security required. Monitoring is key achieving these objectives. It a cornerstone for uninterrupted operation applications’ QoS management. Network tomography uses subset monitoring information, corresponding partial view state, estimate wide-sense performance metrics. In paper, we present machine learning (ML) formulation Tomography. in that its features designed under assumption that: (i) existence certain links not known (e.g., due reasons), (ii) routing dynamic (non-deterministic), i.e., same origin–destination node pair, different route may be selected depending on state links. These assumptions typically modern networks. The can used both additive non-additive We evaluate our proposal using ML algorithms: neural networks (NNs), Gaussian regression linear with interactions. Our simulations indicate has better estimation accuracy compared traditional algebraic or other approaches cannot do take into account two hypotheses. Regarding examined algorithms, differences mainly depend nature