Intelligent Synchronization for Mirrored Game Servers: A Real Case Study

Abstract

Multiplayer Online Games (MOGs) embody intensive applications that require smart solutions able to cope with the high network traffic generated by players, variable latencies, and system failures. To this aim, the anatomy of the game architecture should reflect the possibly wide geographical dispersion of players interacting in a game session. Whereas the use of mirrored game servers has been recognized as a scalable solution to support MOGs, yet, a critical aspect remains that of identifying an efficient synchronization scheme able to responsively guarantee the consistency of the redundant game state. To address this issue, we added intelligence to an optimistic synchronization scheme for mirrored game server architectures: our scheme is able to classify events and, based on their semantics, relax ordering and reliability constraints to gain responsiveness without sacrificing consistency. In this work, we describe the devised scheme and report on an experimental assessment that is based on a real implementation of a mirrored game server architecture, deployed over the Internet. Results definitively show the efficacy of our approach.