Abstract
In this thesis, we reexamine the long discussion on which model is suitable for studying Internet traffic: Poisson or Long-tailed Internet traffic.
Poisson model, adapted from telephone network, has been used since the beginning of World Wide Web, while long-tailed distribution gradually takes over with believable evidence.
Instead of using Superposition of Point Processes to explain why traffic that is not Poisson tends towards Poisson traffic as the load increases, as it is recently claimed in [3], we try to approach this result by another simpler way.