Online gameplay is impacted by the network characteristics of players connected to the same server. Unfortunately, the network characteristics of online game servers are not well-understood, particularly for groups that wish to play together on the same server. As a step towards a remedy, this paper presents analysis of an extensive set of measurements of game servers on the Internet. Over the course of many months, actual Internet game servers were queried simultaneously by twenty-five emulated game clients, with both servers and clients spread out on the Internet. The data provides statistics on the uptime and populations of game servers over a month long period an an in-depth look at the suitability for game servers for multi-player server selection, concentrating on characteristics critical to playability – latency and fairness. Analysis finds most game servers have latencies suitable for third-person and omnipresent games, such as real-time strategy, sports and role-playing games, providing numerous server choices for game players. However, far fewer game servers have the low latencies required for first-person games, such as shooters or race games. In all cases, groups that wish to play together have a greatly reduced set of servers from which to choose because of inherent unfairness in server latencies and server selection is particularly limited as the group size increases. These results hold across different game types and even across different generations of games. The data should be useful for game developers and network researchers that seek to improve game server selection, whether for single or multiple players.
The growth in capability and penetration of broadband access networks to the home has fueled the growth of online games played over the Internet. As this article is being written, it is 4am (EST) on a typical weekday morning and Gamespy Arcade∗ reports more than 310,000 players online playing over 100,000 games! This proliferation in online game players has been matched by an equivalent growth in the variety of online game offerings. The spectrum of online games has shifted from the 1990’s where a few players collaborated or competed on a Local Area Network (LAN) in first-person perspective games such as id’s Doom, to thousands of players interacting over the Internet in a wide variety of games ranging from first-person shooter games and role playing games to real-time strategy and sports games. This escalation in the popularity of online games is also reflected in the correspondingly high number of game servers spread across the globe that support thousands of these players.
Most online games deploy a client-server model of interaction and so could seemingly benefit from network enhancements that support traditional client-server applications. Unfortunately, online games often do not benefit because of significant architectural differences. Traditional clients need specific content from a server and are not interested in alternate versions in the content. In contrast, game clients can often connect to a variety of servers for a different, yet acceptable, gameplay experience with alternate content (maps or players).
Traditional applications care mostly about network throughput, while online games are care mostly about network latency. Traditional applications can benefit from transparent caching and local content access, such as content distribution networks and increased bandwidths. On the other hand, online game players want control over their server selection and especially need specific servers when playing online simultaneously with friends or family.
Many online games allow players to choose from among many servers for online play. For many games, this arrises because users can run their own game servers, allowing clients to connect from anywhere on the Internet. Nearly all popular first-person shooter games (such as Quake, Doom, and Unreal Tournament) allow users to run their own game servers. Similarly, most real-time strategy games (such as Warcraft and Age of Empires) allow users to host a game, thus providing many server choices for clients playing online. And the choice of game server impacts the online experience. Game servers can reach maximum player capacity, require clients to install cheat protection software (such as PunkBuster), or limit access to clients to specific versions or mods. The game map, game configuration and other in-game parameters (such as having friendly-fire enabled for a team-based first-person shooter) can influence a player’s choice to join a particular game server.
Even if all preferential game conditions are met by a game server, network latency will also impact the gameplay experienced. The range of latencies from a game client to all available game servers can be as broad as the range of end-to-end Internet latencies, going from milliseconds for a local game to thousands of milliseconds for a game across the world or over a congested, limited bandwidth network. Previous work has demonstrated that latencies can degrade player performance,1–7 making selection of a fast server important for good online gameplay.
The problem of game server selection can be compounded when multiple players want to play online simultaneously on the same game server. This can arise when friends and family arrange for an on-line gaming session or when more formalized teams of players (typically called clans) compete against other teams during a pre-arranged match. Given the increasingly connected nature of the world, such players are increasingly likely to be geographically separated. A server that is fast for one player may be slow for the next player and vice versa.
Support for finding a game server that performs acceptably for all players that want to play together remains an open research issue.
In order to improve game server selection, both for single and multiple players and for the increasingly diverse set of online games they play, there is a need for a better understanding of the network characteristics of current game servers. Understanding the availability and performance for existing game servers will provide insights as to whether there need to be alternate means for server deployment. Measuring game server performance from game clients simultaneously running on different Internet nodes will enable assessment of support for both single and multiple players, and can be extrapolated to assess support for a wide-range of game genres.
Some related work8–10 has looked at improving server selection for the current client-server online games, but has focused almost exclusively on helping single players without considering support for server selection for simultaneous players. Other related work11, 12 proposes alternate structures for organizing game servers to provide for better online gameplay, but are not immediately useful for today’s predominantly client-server online game environments. Related work in third party architectures13, 14 shows promise in improving server selection for games and other client-server applications, but does not work without explicit buy-in by game developers. Other research has analyzed game traffic from the point of view of a server15, 16 or a client,17–20 but does not specifically consider traffic patterns or performance criteria for a group of servers and a group of clients.
In an effort to provide a better understanding of current game server selection, this paper provides a network characterization of actual game servers on the Internet. An extensive set of measurements were gathered over the course of several months, involving 25 geographically dispersed game clients and 60 geographically dispersed game servers from 3 different game types. The actual Internet game servers were queried simultaneously by the game clients, providing statistics on the uptime and populations of game servers and an in-depth look at the suitability for game servers to support single and multiple player server selection for a range of game genres.
Analysis of the data finds most game servers have latencies suitable for third-person and omnipresent games, such as real-time strategy, sports and role-playing games, providing numerous server choices for game players, whether selecting a server alone or simultaneously with friends and family. However, far fewer game servers have the low latencies required for first-person games, such as shooters or race games. Groups that wish to play together have a reduced set of servers from which to choose because of inherent unfairness in server latencies. These results hold across different game types and even across different generations of games.