Degree Name

Master of Information and Communication Technology - Research


School of Information Technology and Computer Science - Faculty of Informatics


The distribution of game servers has received serious attention in recent years. A multi-server structure may distribute the load of the server, as well as improve the fairness of games played over large geographical distances without sacrificing delay. However, while distributing the game authority, the system takes risk of game state inconsistency between servers. Mechanisms are presented to provide synchronization between servers. However, user response to these mechanisms for maintaining server consistency in different game types is largely unknown. This thesis aims to study the user responses to two of the important synchronization techniques, Local Lag and Timewarp, within a real network game. To do this we developed a platform based on a common First Person Shooter (FPS) game called Quake III Arena. The platform is able to simulate various distributed server systems. With the platform we examines the impacts of using local lag/timewarp upon computer-controlled players (bots) performance in distributed network game. According to our experimental results it is probable to gain the optimal point of bots performance with constant network lag while using Local Lag and Timewarp together. To discuss the probable difference between human players and bots, and to examine the practicability of using the bot system to simulate human players' behaviors, we perform experiments with the participation of human players, and compare the results between bots/human players performance while under the same network conditions. The two groups of results shows significant similarities. Finally, in order to analyze and arrange the synchronization of distributed system more efficiently, we present Virtual Server Shifting Theory. We first demonstrate the - i - practicability of applying this theory within the distributed server system based on the mirrored server system, and then generalize the conclusions to more complex distributed server network topologies.

02Whole.pdf (2032 kB)