Performance Modeling For Game Engine Based On Queuing Network

With advances in computer network and technology, online games gradually become important forms of entertainment. In online games, the server system need to support hundreds of simultaneous users distributed all around. Performance of such system has become one of the key issues because of the heavy processing task. On the other hand, game engine is the abstraction of infrastructure of online game system. Its quality determines the system's stability, scalability, performance and so on.This dissertation focuses on a online game engine architecture based on area partition. An analytic model of the engine is presented .The workload influenced by number of users, number of servers, size of game world, size of AOI and the traffic intensity is studied with the arrival rate of the network packet as the performance metric. It can be further analyzed that with the load increasing, whether the new performance requirements can be meet through expanding the system. Analysis result shows that the processing capacity can be enhanced by adding more servers to the system. Thus, the system shows good scalability.The server delay is an important performance metric of online games. A queuing network model is presented and the composition of server delay is studied. Real and accurate parameter is measured by simulate experiment using actual game server engine. According to the model, the server delay influenced by number of servers, processing capacity of server, number of users, the traffic intensity, etc. is studied. The result shows that the size of AOI and traffic intensity has greater influence to server delay and the delay can be significantly decreased by improving the processing capacity of game servers.