Quantification of Capture Effect in Near-Far Scenarios Month 1998 doc.: IEEE 802.11-98/xxx May 2001May 2001 Quantification of Capture Effect in Near-Far Scenarios Greg Chesson (greg@atheros.com) Aman Singla (aman@atheros.com) Atheros Communications, Inc A. Singla, Atheros Communications, Inc John Doe, His Company
Known problems OS resource mis-management Protocol Stack adaptation May 2001May 2001 Known problems OS resource mis-management one application can “capture” OS resources: buffers, ifq, … Protocol Stack adaptation TCP treats packet loss as congestion TCP adapts by reducing offered load (window size, slow-start) PHY rate adaptation NIC or adaptive device driver reduces PHY rate to compensate for packet loss Near-far effects near station may block/drown transmissions from far stations A. Singla, Atheros Communications, Inc
Objectives Reproduce near-far effect in simulation May 2001May 2001 Objectives Reproduce near-far effect in simulation Quantify and analyze the magnitude of near-far effect A. Singla, Atheros Communications, Inc
May 2001May 2001 Scenario 36Mb/s 802.11a PHY Near/Far station determined on basis of the relative strengths of the signals originating from the station as received at the AP All stations are visible to each other Collisions between a far station and a near station resolve in favor of the near station Simulation parameters adjusted to guarantee capture by near station(s) Near Stations AP 1500 byte TCP/UDP Streams Far Station A. Singla, Atheros Communications, Inc
UDP Results May 2001May 2001 A. Singla, Atheros Communications, Inc Near Stream Far Stream A. Singla, Atheros Communications, Inc
TCP Results May 2001May 2001 A. Singla, Atheros Communications, Inc Near Stream Far Stream A. Singla, Atheros Communications, Inc
May 2001May 2001 Conclusions Far station loses “constant” UDP bandwidth to near station(s) percentage increases with number of near stations Less degradation for TCP traffic data sink has a moderating effect less “loss” with increasing number of near stations Limited experimental scope no correlation with range/signal data limited number of experiments limited topology A. Singla, Atheros Communications, Inc