Academics from the University of Lincoln, UK, have investigated game server performance metrics, and how they may best be used to explore the behaviour of multicore server processes.
The work focuses on small-scale stand-alone game servers, running on standard consumer PC architectures, which support a large proportion of online game play. These servers are highly sophisticated and complex software applications; understanding their behaviour goes hand-in-hand with performance optimisation.
The research was conducted by James Munro, a research student at the School of Computer Science, along with Dr Patrick Dickinson and Dr Kofi Appiah.
He said: “With some online games there are dedicated servers but many are run on standard consumer equipment, such as people’s own computers at home. Player experience is often compromised due to the speed the server can process the information, making server optimisation an on-going challenge for developers.
“It is difficult to critically compare previous studies due to the inconsistent, ad-hoc metrics used to analyse performance with little consideration of how to interpret results. Our research centred on finding the best metrics and how you can best compare and therefore optimise simple server architectures.”
The team suggest that people who want to do comparisons of these types of games on these architectures should consider using a specific set of metrics they proposed, to establish a common set of comparable parameters.
They collated metrics used in these studies, proposed some new ones, and through a series of experiments examined their inter-relationships and how they may best be used to investigate the behaviour of multicore server processes.
Dr Dickinson said: “We concluded our experiments by using our suite of metrics to show how the relationship between game mechanics and performance can be quantified and used to inform game design. We see the continued development of empirical analysis based on server metrics as an area which is not only attracting academic interest, but also has the potential to deliver new and useful tools to industry.”
The paper ‘Investigating informative performance metrics for a multicore game world server’ is published in the journal of Entertainment Computing Volume 5, Issue 1, January 2014 http://www.sciencedirect.com/science/article/pii/S1875952113000165