On the Geographic Distribution of On-line Game Servers and Players

This is another characterization paper I've read as a reference for one paper I've just submitted about online game streaming. It was written by researchers from Oregon Health and Science University. They have also authored another paper about online gaming already summarized here. The idea of this particular paper is very simple, characterizing how players and servers are distributed over the world. They argue that this kind of analysis is relevant to improve player experience, specially in terms of provisioning server resources  globally.

Even not being very much into gaming, I'm aware that lagging really damages player experience. In order to reduce lags, it is interesting to place servers close to players in the network. As gaming services become a high profiting business, companies put a lot of effort into resource allocation, including server placement. The authors have an hypothesis that players are usually close to the servers as a natural consequence of lagging.

The dataset used in this analysis are counter-strike server traces (for player distribution) and lists of game servers returned to game clients for different games (for server distribution). Geographic locations were generated by a commercial geographic mapping tool.

I really liked the way game and player server distributions can be visualized in the paper. The authors use longitude histograms, which show the volume of players/servers in a given meridian. I've done a similar analysis to show how streamers are distributed on Twitch.Tv for my paper on game streaming. In general, game servers concentrated mostly in the U.S., specially in the west cost. There are also many servers in Europe and Asia. In terms of latitude, the great majority of the servers are in the northern hemisphere. Players follow a similar geographic distribution. However, the authors found that 50% of the accesses to the server com from outside of North America. Several hypotheses for this result are enumerated, including disparities between geographic and network position, malfunctioning of server selection mechanisms, and shortage of servers overseas.

Another interesting result presented in the paper is that the geographic distribution of players is correlated to the time of the day, which motivates the shift of servers according to demand. Moreover, when compared to the geographic distribution of accesses to a department website, players are closer to the game server. accesses to a web site that provides game statistics and forums about games was also found to be close to the server.

This paper does exactly what it aims to do and it is not very audacious. The analysis are intuitive and the visualizations are good.

Link: http://www.thefengs.com/wuchang/work/cstrike/netgames03_geo.pdf