An exponential arrival model for Half-Life 2 server discovery traffic

Online multiplayer first person shooter (FPS) clients typically discover public game servers through a two-step process. A master server is queried for a list of current game servers, who are then sequentially probed (creating a 24-hour cycle of background noise for FPS game servers). Using 3 million probe packets received by a Half- Life 2 server in February and April 2006, we show arrivals exhibit exponentially distributed, uncorrelated inter-probe intervals during both busiest and least-busy hours of a day. We further show a modified Laplace curve is a reasonable per-hour estimator of for the exponentially distributed probe arrivals. I