Using local lag and timewarp to improve performance for real life multi-player online games

  • Authors:
  • Dingliang Liang;Paul Boustead

  • Affiliations:
  • University of Wollongong, Australia;University of Wollongong, Australia

  • Venue:
  • NetGames '06 Proceedings of 5th ACM SIGCOMM workshop on Network and system support for games
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

When developing distributed server games it is important to consider consistency of decisions made at different servers as well as the response time experienced by the players. This paper examines two approaches that are commonly used to improve the consistency of distributed servers: local lag and timewarp. Local lag can be used to remove inconsistencies, however it can significantly increase the response time experienced by players. Timewarp on the other hand allows inconsistencies to exist, for a short time, before rolling back time and correcting the decisions in the servers. This paper aims to produce a platform to experimentally evaluate the effect of each of these on players. To do this we have developed a distributed version of Quake III that includes both timewarp and rollbacks. In this paper we present the first set of results obtained with this test-bed. These results examine the performance, based upon the score in the game, of computer-controlled players, or "bots". We present results on how varying both local-lag, and network lag (with and without timewarp) effect "bot" performance. With the results it shows that the combinative use of local lag and timewarp could achieve a similar effect of half reducing the network lag. We plan to use the test-bed to test the effect on real players at a later stage.