THE INFLUENCE OF NETWORK TOPOLOGICAL MODELS ON THE PREDICTION OF END-TO-END LOSS PROBABILITIES

In last years new procedures to generate Internet-like topologies have been proposed, stemming from the evidence of a power-law behavior of real networks. Realistic networks generators are necessary for simulation and performance evaluation of data communication systems. Assuming a simple loss model for the links, the present investigation shows that the adoption of a specific topological model for the Internet graph (here, the Waxman or the Barabási-Albert model) may affect, to a certain extent, the estimates of the loss probabilities for end-to-end communications obtained by simulation (e.g. via NS-2 package).