This paper discusses issues in providing immersive audio communication to massively multi-player online games (MMOG). We focus on network and server architectures for creating immersive audio scenes for game clients. We examine advantages and limitations of several architectures, including central server, distributed servers, and peer-to-peer architectures. We focus on a dynamic centralized server architecture which is suitable for the current MMOG in terms of sealability at end users and resource efficiency. We propose algorithms to optimize the performance of a centralized server. Finally, a simulation study is carried out to evaluate the performance of our proposed centralized server architecture and provide some comparisons with others in term of delays and network resource usage.