Further Optimising Online FPS Game Server Discovery

This paper proposes several improvements to a proposed game server discovery algorithm based on clustering servers by origin Autonomous System (AS). Following a brief overview of current game server discovery processes, the algorithms on which this paper is based are outlined. Using Valve's Counterstrike:Source game and datasets employed in the previous algorithm's paper, several improvements to the subclustering and calibration algorithms are explored. This includes sub-clustering based on a dynamic choice of network prefix and alternative definitions for a function to choose the number of calibration probes. The suggested enhancements allow a significant reduction in the number of calibration probes required while maintaining reasonable reordering of servers. Subsequently, appreciable reductions in the time and network traffic required over the existing proposed algorithms are observed. KeywordsServer discovery, search optimisation, latency estimation