Rate control methodologies that are currently available in IEEE 802.11 network cards seriously underutilize network resources and, in addition, per-second throughputs suffer from high variability. In this paper, we introduce an algorithm, H-RCA, that overcomes these shortcomings, giving substantially higher, and less variable, throughput. The approach solely uses information already available at the driver-level to function and can be implemented on 802.11e commodity hardware. H-RCA's design objective is to minimize the average time each packet spends on the medium (including retries) in order to maximize total network throughput. It uses a development of a recently proposed estimation scheme to distinguish transmission failures due to collisions from those caused by channel noise. It employs an estimate of the packet loss ratio due to noise in assessing whether it is appropriate to change rate. We demonstrate experimentally that packet loss ratio is not necessarily a monotonic increasing function of rate; this is accounted for in H-RCA's design. As H-RCA statistically separates noise losses from those caused by collisions, ns-2 simulations show that it is robust to changing environments. H-RCA does not require specific hardware support nor any change to the IEEE 802.11 protocol. This point is substantiated with results from an experimental implementation.

The robustness to noise of the 802.11b/g 5.5 Mb/s and 11 Mb/s rates must be investigated experimentally as they cannot be predicted theoretically. In this paper we report on detailed outdoor and indoor measurements that lead us to the surprising conclusion that the 11 Mb/s 802.11g rate experiences fewer packet losses than the 6 Mb/s 802.11g rate at any given (symbol) SNR. This occurs due to the combination of modulation and physical layer coding schemes used by these rates and has serious implications for rate control algorithms. The practical implications of this, factoring in the interaction between packet loss and 802.11 MAC retries, is that 6 Mb/s is effectively redundant as a packet transmission rate if the 11 Mb/s rate is available.

We identify common hypotheses on which a large number of distinct mathematical models of WLANs employing IEEE 802.11 are founded. Using data from an experimental test bed and packet-level ns-2 simulations, we investigate the veracity of these hypotheses. We demonstrate that several of these assumptions are inaccurate and/or inappropriate. We consider hypotheses used in the modeling of saturated and unsaturated 802.11 infrastructure mode networks, saturated 802.11e networks, and saturated and unsaturated 802.11s mesh networks. In infrastructure mode networks, we find that even for small numbers of stations, common hypotheses hold true for saturated stations and also for unsaturated stations with small buffers. However, despite their widespread adoption, common assumptions used to incorporate station buffers are erroneous. This raises questions about the predictive power of all models based on these hypotheses. For saturated 802.11e models that treat differences in arbitration interframe space (AIFS), we find that the two fundamental hypotheses are reasonable. For 802.11s mesh networks, we find that assumptions are appropriate only if stations are lightly loaded and are highly inappropriate if they are saturated. In identifying these flawed suppositions, this work identifies areas where mathematical models need to be revisited and revised if they are to be used with confidence by protocol designers and WLAN network planners.

As WLANs employing IEEE 802.11 have become pervasive, many analytic models for predicting their performance have been developed in recent years. Due to the complicated nature of the 802.11 MAC operation, approximations must be made to enable tractable mathematical models. In this article, through simulation we investigate the veracity of the approximations shared by many models that have been developed starting with the fundamental hypotheses in Bianchipsilas (1998 and 2000) seminal papers. We find that even for small numbers of station these assumptions that hold true for saturated stations (those that always have a packet to send) and for unsaturated stations with small buffers. However, despite their widespread adoption, we find that the commonly adopted assumptions that are used to incorporate station buffers are not appropriate. This raises questions about the predictive power of models based on these hypotheses.