Analysis of multi-player online game traffic based on self-similarity

Abstract

When provisioning network resources, ISPs have to tackle with the difficulties imposed by the traffic characteristics and demands of the users. Multi-player online games (MPOG) are constituting a large portion of the Internet traffic. MPOG players demand high bandwidth and low delay due to high interaction in these games. Moreover, they generate bursty traffic. In this paper, we analyze the traffic characteristics of a popular MPOG type game; Call of Duty 2 version 1.0 (CoD2). We use several game sessions to extract packet size and inter-arrival time distributions. In addition to these we investigate the self-similarity of the CoD2 traffic. We observe two different behavior at server and client-side, as expected. For the server-side, traffic characteristics vary with the number of players, but the traffic is long-range dependent most of the time. There are several intervals when the traffic process is non-stationary where the analysis of self-similarity becomes trivial. For the client-side, the traffic shows little variations however there are still some cases where the analysis of self-similarity becomes intricate because the traffic process reveals deviation from stationarity.