User behaviour based modeling of network traffic for multiplayer role playing games Mirko Suznjevic University of Zagreb, Faculty of Electrical Engineering and Computing Unska 3, Zagreb, Croatia mirko.suznjevic@fer.hr 2 nd TMA PhD school Napoli, June 2011
Outline Introduction Problem Proposed approach Classification of the virtual world state Traffic modeling Player behaviour modeling Implementation Conclusion Napoli, June 2011 2 /15
Introduction - MMORPGs Massively Multiplayer Online Role-playing Games (MMORPGs): Persistent virtual worlds Players represented through avatars Many possible interactions Popularity of MMORPGs is increasing Players motivation? Game developers and publishers 317 published and 83 in development Use case World of Warcraft Napoli, June 2011 3 /15
Introduction - MMORPGs Number of active MMORPG subscriptions (taken from www.mmodata.net) Napoli, June 2011 4 /15
Bandwidth [PB per month] Introduction - MMORPGs 250 200 150 100 Fixed Mobile 50 0 2010 2011 2012 2013 2014 2015 Time [years] Cisco Visual Networking Index:Forecast and Methodology, 2010 2015 Global Consumer Internet Gaming Traffic Growth Napoli, June 2011 5 /15
Problem I Highly variable bandwidth usage Same application! Napoli, June 2011 6 /15
Problem II Server traffic patterns M. Kihl, A. Aurelius, and C. Lagerstedt, Analysis of World of Warcraft traffic patterns and user behavior, in International Congress on Ultra Modern Telecommunications and Control Systems and Workshops, 2010, pp. 218 223. Napoli, June 2011 7 /15
Proposed approach Classification of the virtual world state Defined 5 player action classes Independent of a specific game Validation of established classes Network traffic measurement and analysis Traffic modeling for every class Implementation of models in D-ITG Player behaviour modeling Markov chain for a single player Poisson process for session arrivals Implementation and validation Napoli, June 2011 8 /15
Classification of the virtual world state M. Suznjevic, O. Dobrijevic, and M. Matijasevic, MMORPG player actions: Network performance, session patterns and latency requirements analysis. Multimedia Tools and Applications, 45(1-3):191-241, 2009. Napoli, June 2011 9 /15
Number of packets (pps) Bandwidth usage (kbps) Validating established player action classes 25 9 20 15 10 5 8 7 6 5 4 3 2 1 Questing Trading PvP combat Raiding Dungeons 0 0 Packet rate Bandwidth usage Napoli, June 2011 10 /15
Traffic modeling Traffic highly erratically distributed - modeled by composition of several distributions Client APDU sizes: Modeled as discrete values (14B and 35B are the most common) Client IATs Discrete values 0 and 500ms + Weibull distribution Server APDU sizes: Lognormal distribution or split consisting of two Weibull distributions Server IATs Several discrete values (44ms, 200ms, 218ms, 328ms) + Lognormal or Weibull Napoli, June 2011 11 /15
Behaviour modeling Measurements passive measurement by using a software game add-on, deployed on players' computers Detailed information, but low player count (104) Session consists out of several action specific segments Hour specific modeling of segment duration (Weibull, Smallest Extreme, Largest Extreme) Switching between segments modeled by a hour specific Markov chains Arrival and departure processes extracted from existing dataset (Y.-T. Lee, K.-T. Chen, Y.-M. Cheng, and C.-L. Lei, World of Warcraft avatar history dataset, in Proceedings of the second annual ACM conference on Multimedia systems, 2011, pp. 123 128) Napoli, June 2011 12 /15
Behaviour - traffic relationship Napoli, June 2011 13 /15
Implementation Work in progress Simulator of player behaviour Simulation of the traffic load on a single WoW server Linux Container which runs D-ITG as represents a single user Napoli, June 2011 14 /15
Conclusion Established player action classes Explained the relationship between the traffic patterns and player behaviour Modeled the player behaviour Modeled the MMORPG traffic per action class Future work: Investigation of transformational schemes for modeling the traffic Validation of behaviour patterns based on other available player behaviour logs Napoli, June 2011 15 /15