Abstract

In this paper we investigate optimal load balancing
strategies for a scalable parallel game server architecture.
Our work develops from an existing multi-threaded implementation
of the QuakeWorld game server: we investigate
the comparative effectiveness of different load-balancing
algorithms, and determine how different metrics can be
used to analyse performance. We find that achieving optimal
QuakeWorld server performance is a trade-off between
consistently achieving an even workload distribution whilst
reducing intra-frame wait time, and that a combined set of
metrics is required to fully understand how load balancing
affects server performance.