Wide Area Traac Modeling: from Poisson to Self-similarity

Telephone networks can be well modeled using an exponential distribution because they show a relatively homogeneous and predictable behavior over large time scales. In contrast to these circuit-switched networks the Internet has very diierent characteristics. At rst people tried to create simulations of this global network with the same tools that were used before, but soon realized that its traac was much more variable over time. Therefore , a new model had to be developed that could cope with this special behavior. Other distributions are now used in place of Poisson which more adequately reeect the packet-switched, highly dynamic Internet.