Abstract
Video games have proven to be an interesting platform for computer scientists, as many games demand the latest technology, fast response times and effective utilization of hardware. Video games have been used both as a topic of and a tool for computer science (CS). Finding the right games to perform experiments on is however difficult. An important reason is the lack of suitable games for research. Open source games are attractive candidates as their availability and openness is crucial to provide reproducible research. Because researchers lack access to source code of commercial games, some create their own smaller prototype games to test their ideas without performing tests in large-scale productions. This decreases the practical applicability of their conclusion. The first major contribution of the thesis is a comparative study of available open source games. A survey shows a list of demands that can be used to evaluate if a game is applicable for academic use. The study unveils that no open source game projects of commercial quality are available. Still, some open source games seem useful for research. These can be used for the implementation, testing, and verification of scientific concepts. The comparative study suggests that Doom 3, OpenArena (Quake III Arena), and PlaneShift are some of the best candidates available as of today. None of the suggested games are ideal for testing server or network specific concepts, as none provides tools for simulating user-generated load. This is needed to perform a load test on a server, to evaluate the effect of the implemented concepts. Network traffic and system load can be generated through the implementation of clients controlled by artificial intelligence (AI) that simulates real players. They should produce real network traffic and server load that is similar to traffic produced by a real human player. These specialized clients are often referred to as virtual clients. The second major contribution of this thesis describes the process of converting the open source video game Quake III Arena, into a user- friendly load generation tool for investigating game system scalability. This has been done through the implementation of virtual clients, providing researches with an automated procedure to evaluate network and server performance under various loads. The tool proves to both generate authentic traffic and server load.