Analyze and improve the QoS of multiplayer games over 802.11 at the MAC layer

1. Quality of Service of Online Multiplayer Games over Wireless
LAN( IEEE 802.11)
Ashwin Murali
Masters in Telecommunications
A. James Clark School of Engineering
Abstract Method Results The WME parameters were varied for VI,VO queues.
Background traffic UDP: 2500pps, 1400 bytes AIFS CWmin TXOP limit
Online gaming is one of the most rapidly increasing sources of traffic in Setup a wireless testbed consisting of 5 PC’s and 4 soekris boxes VI 7 4 2048
the modern internet. VO 1 2 3008
Operating System: linux Interarrival time with traffic
The poster contains the details of the experiments performed to evaluate Interarrival time with priority queuing
the performance of FPS games over WLAN in the presence of UDP and 4 boxes to AP
TPlink wireless card
TCP background traffic. 4 boxes to AP
Madwifi driver
A wireless testbed of 5 PC’s and soekris boxes was setup to evaluate the
performance of multiplayer games over WLAN. WLAN: 802.11g
Quake 4 by ID software which is one of the most popular FPS games and Traffic generation
which also has a Linux install version was used for the experiments.
MGEN for UDP traffic, Iperf for TCP traffic
Parameters such as packets per second, inter-arrival times and queue size Minor lag at C1, C2
plots were analyzed. Game c2
server No lag was observed at C1, C2
The results were compared with those from previous such experiments by AP to 4 boxes
Testbed
other research groups ( Swinburne University of Technology, Simulating server AP
Online Network Games(SONG) database). 15 UDP flows were possible as background traffic to the AP before lag
was observed.
Techniques such as queue prioritizing were employed to reduce the delay
Tests were performed with TCP as background traffic.
enhance the game play.
This did not affect the game play as UDP is not TCP friendly.
c1
soekris Conclusion
Lag noticed on C1, C2
Introduction The lag observed at the clients with background UDP
Interarrival time without
Made changes to AIFS, CWmin, TXOP traffic is mainly due to congestion at the AP.
Data collection background traffic
parameters.
Different types of multiplayer games exist.
TCPdump was used to
This did not reduce the lag.
First Person Shooter, Real Time Strategy, Simulation are most popular. collect data Prioritizing the queues is an option to reduce the lag and
enhance the game play.
FPS: Quake, Unreal Tournament MATLAB/GNUplot used Tests with the wireless driver modified
RTS: Age of Empires, Warcraft to plot the graphs However this does not eliminate the lag completely.
The wireless driver was modified to include different queues for
Different games use different transport layer protocols: Game traffic: gaming and background traffic.
FPS games: UDP Server to client::~200 Variation of WME parameters such as AIFN, CWmin and
RTS games: TCP bytes The traffic was differentiated based on packet size. TXOP limit along with priority queuing should also be
Client to server:50-80
bytes Highest priority AC_VO queue was used for game traffic.
considered.
Netwrok Architecture:
FPS games use a client-server model as opposed to a peer tp peer model
used by RTS games. Interarrival time For background traffic initially AC_BE queue was used.
At server: 15ms
Game times in FPS games are usually short. At client: 85-95ms However the WME parameters of the AC_BE queue at the AP
could not be changed.
FPS games are most sensitive to delay( <150ms). Packets per second
At servr: 14 AC_VI queue was used instead.
Acknowledgements
The varied performance of the IEEE 802.11 standard has tended to make At client: 14 Research described in the poster was supervised by Dr.
Over LAN Lag was observed on both C1 and C2.
WLANs unsuitable for FPS games. Douglas Leith, Director, Hamilton Institute, Ireland and
Dr. David Malone, Hamilton Institute and supported by
the Science Foundation Ireland.