Poster abstract: measuring multi-parameter conflict graphs for 802.11 networks

Dropping prices and increasing demand for a mobile workforce have caused a proliferation of wireless LANs in modern enterprises. In an attempt to increase wireless capacity and provide complete coverage, enterprises are turning to dense deployments of access points (APs). Unfortunately, APs on the same channel that ‘share the air’ are interference-limited, due to hidden and exposed terminals. Thus, increasing AP density can, paradoxically, reduce aggregate throughput. Proposals aiming to mitigate RF interference in 802.11 networks, such as dynamic frequency selection, power control, and traffic scheduling [1] assume the existence of a conflict graph (CG) as input. A conflict graph is a data structure that encodes interference information between links in the network. Specifically, for each link, it lists the set of other links which could cause interference either at the transmitter or receiver of the link. Note that the existence of a conflict edge between two links in a CG is dependent on factors such as the power level of the interferer and the data rate used on the link. Thus, what is needed is to determine for each link, the conditions under which all other potentially interfering transmitters could cause interference on that link. Measuring such configurations using today’s state of the art approaches [4] turns out to be highly unscalable. Our goal, therefore, is to find efficient ways of measuring conflict that incorporates all relevant effects (i.e. build a multi-parameter CG), while still being computationally efficient. We conduct a measurement study to gain a deeper understanding of how data rate, transmit power and temporal changes affect the conflict that exists between two links in a network. The insight we obtain allows us to construct a multi-parameter CG without measuring the entire state space of possible configurations. Contributions: The following is a summary of some of our findings: