Traffic load distribution in massively multiplayer online games

Abstract

Online games have been improved dramatically these days. They possess tens of thousands of concurrent players. It is expected to provide multimedia streaming in online game area like other kinds of computer or video games. Game titles currently serviced exploit client-server architectures and scalability problem due to single server has always been a hot issue in this type of entertainment applications. It will give the same limitation when developers try to realize streaming functionality. We consider a federated peer-to-peer architecture, a hybrid of client-server model and pure peer-to-peer model is one of feasible options to enable the scenario. In this model, reflectors are put between clients and control server to forward packets from each client. They can disseminate stream traffic volume so that prevent single server overload problem in conventional client-server model. However, a reflector can be overwhelmed when client distribution is skewed. Too many clients connect to one reflector and it creates network traffic bottleneck. We suggest a traffic distribution model and test its applicability. Simulation result indicates that crowd size affects process time and redistribution cost. Suggested model succeeds to reduce peak traffic of overloaded reflector to less than twice of average traffic size.