Improving fairness between close Wi-Fi access points

We consider some Wi-Fi access points (AP) communicating with some stations while being in the carrier sense area of neighbor AP. This corresponds to usual scenarios in local area networks, set-top boxes in houses, intelligent transportation systems, etc. First, using the network simulator ns-3, we observe interesting phenomena, depending on the number of pairs and its parity, confirming previous experiments and models of the literature. Short chains give high disparity in throughput. Moreover, adding a pair (i.e., an AP) can drastically change the throughput of a far ones. We also point out a notable asymptotic behavior. We show that the same phenomena appear while varying the number of stations per AP, the sending rate or the chain geometry. Second, we provide a new analytic model by considering the probability for a station to send data while its neighbors are waiting. This powerful model leads to a non-linear system of equations matching very well the N's-3 simulations. It is scalable and permits to study accurately both short and very large chains. In particular, we show interesting properties related to the fairness. Third, we apply our results to some practical cases. Thanks to our model, we are able to compute the optimal settings of the stations for ensuring the fairness in the scenarios listed above.