Characterization and modeling of campus-level IP network traffic

The behavior of IP based network traffic has been the subject of several works. In this paper, we compare two modeling approaches and their ability to capture two basic traffic features: packet size distribution and interarrival times. Data traces were measured at the border gateway of the campus network of the University of Pavia at different times of the day. The two modeling schemes use respectively well established hidden Markov model (HMM) techniques and a stochastic engine derived from chaotic attractors.