background image
Empirical Performance Evaluation of the Zereal
Massively Multiplayer Online Game Simulator
Amund Tveit
Division of Intelligent Systems
Department of Computer and Information Science,
Norwegian University of Science and Technology,
N-7491 Trondheim, Norway
amundt@idi.ntnu.no
Abstract. This paper presents a factorial experimental design for test-
ing which factors controlling the scalability of the Zereal Massively Mul-
tiplayer Online Game simulator. The analysis show that the factors ex-
plain approximately 97% of the model (measured in square of errors),
which is surprisingly high since the simulation is stochastic.
1
Introduction
In this paper we will investigate the scalability of the Zereal Massively Multi-
player Online Game Simulator, and how much of the response variable that can
be explained by the factors (e.g. simulator parameters).
The selected experimental method is factorial design since it enables the
measurement of both single impact and interaction effects of the parameters
used to control Zereal. Selected outcome measurement is wallclock time for a
fixed number of simulation steps. More details about the statistical analysis and
data can be found in appendix C
1.1
Purpose of Zereal
The main purpose of the Zereal simulator is to provide a scalable (research)
testbench for testing models of players, monster intelligence, and various data
analysis and data mining approaches to massively multiplayer online games [2].
The alternative way of performing such research would require tight cooperation
with major massively multiplayer game vendors like Microsoft, Sony, Electronic
Arts, which requires extensive negotiation rounds since (real) player data is one
of their greatest assets.
1.2
Goal of analysis
Get a detailed overview of how (and how much) various parameters (i.e. factors)
affect the runtime of the simulation.
92
Performance of the Zereal MMOG Simulator

<< - < - > - >>