A packet-level traffic model of Starcraft

Starcraft is a popular real time strategy game that uses a peer-to-peer network communication model. In this paper we analyze its traffic and we provide a statistical characterization at packet-level obtained by varying the number of players. We examine the time dynamics between individual packets within a game session as well as the packet sizes. Also, we provide analytical models approximating the empirical distributions found, we study properties of the tails and of the auto-correlation function, and we investigate the presence of self-similarity. The results obtained show how traffic generated by such game has different characteristics from the traffic prevailing on the Internet in past years.