The Snowball Effect: Detailing Performance Anomalies of 802.11 Rate Adaptation

Current rate adaptation schemes for 802.11 exhibit sudden and severe drops in throughput in real application scenarios. Although there have been several propositions to remedy such rate adaptation failures, there has not been a thorough analysis of the causes that lead to such a "snowball effect." This paper provides an analysis of the factors that lead to the poor performance of current rate adaptation schemes in real environments. We show that current rate adaptation schemes fail because they do not differentiate between poor channel conditions and collisions as the source of transmission failures, and consequently invoke improper responses that cascade to dramatic throughput degradation. We support the analysis through experimentation with real data from a wireless video surveillance application, and provide recommendations for the next generation of WiFi rate control schemes.