Characterization of 802.11n wireless LAN performance via testbed measurements and statistical analysis

The 802.11n standard introduces a number of new MAC and PHY features to achieve high throughput and reliability. We conduct a comprehensive characterization of 802.11n performance with respect to its constituent features across a wide variety of scenarios with the aid of 802.11n wireless LAN testbed based measurements and statistical techniques including regression analysis. Our results show that different 802.11n features are interdependent when optimizing performance metrics such as throughput; the nature of interdependence as well as their relative impact are scenario dependent. We show the feasibility of online sender-side interference type detection, a key part of identifying the operational scenario for comprehensive 802.11n link adaptation, via a supervised machine learning based classifier. Finally, we highlight the unfairness problem of 802.11n networks that is linked to the frame aggregation feature.