Novel metrics and experimentation insights for dynamic frequency selection in wireless LANs

Abstract

The rapidly increasing popularity of IEEE 802.11 WLANs has created unprecedented levels of congestion in the unlicensed frequency bands, especially in densely populated urban areas. Performance experienced by end-users in such deployments is significantly degraded due to contention and interference among adjacent cells. In this paper, we develop novel metrics and insights that we use for dynamic frequency selection, incorporating the various features that affect interference. The proposed scheme features a novel client feedback mechanism, which enables nodes of the cell, as well as nodes belonging to different cells, to contribute to interference measurements. Furthermore, we incorporate a traffic monitoring scheme that makes the system aware of prevailing traffic conditions. We design a distributed protocol, through which messages containing the information above are passed by the stations to the access-points, where the frequency selection is performed in a dynamic form. The proposed algorithm is implemented in the Mad-WiFi open source driver and is validated through extensive testbed experiments in both an indoor RF-Isolated environment, as well as in a interference-rich, large-scale wireless testbed. Results obtained under a wide range of settings, indicate that our algorithm improves total network throughput, up to a factor of 7.5, compared to state-of-the-art static approaches. © 2011 ACM.