On the fidelity of IEEE 802.11 commercial cards

The IEEE 802.11 DCF protocol is known to be fair in terms of long-term resource repartition among the contending stations. However, when considering real scenarios, where commercial 802.11 cards interact, very unpredictable as well as sometimes surprising behaviors emerge. Motivation of this paper is to investigate the reasons of the very evident disagreement between the theoretical IEEE 802.11 DCF protocol models and its practical implementations. In particular, we try to characterize the card behavior not only in terms of perceived throughput, but also in terms of low-level channel access operations. In fact, the simple throughput analysis does not allow to identify what affecting parameters, both in terms of transceivers architectures and MAC layer deployments, determine the performance differentiation among the cards. To this purpose, we implemented a tunable DCF network card, in which all MAC parameters are programmable and all the baseband signals are available, and we used this card as a probe instrument. We registered the low-level access operations of commercial cards in terms of access times revealed by the carrier sense function of our probe card. By comparing these times, we surprisingly proved that the most evident performance differences are not due to PHY layer issues, but to the MAC implementations, which often seem to do not respect the standard specifications.