Cell-Level Modeling of IEEE 802.11 WLANs

TL;DR

This paper introduces a scalable cell-level analytical model for multi-cell IEEE 802.11 WLANs that accurately predicts throughput and delay, especially under conditions where hidden nodes are present or the PBD condition holds.

Contribution

It develops a novel cell-level model incorporating network topology effects and an effective service rate approximation, improving throughput and delay predictions over previous models.

Findings

Accurate throughput predictions for saturated nodes and TCP downloads.

Effective delay approximations using the proposed model.

Model performs well even with hidden nodes present.

Abstract

We develop a scalable \textit{cell-level} analytical model for multi-cell infrastructure IEEE 802.11 WLANs under a so-called Pairwise Binary Dependence (PBD) condition. The PBD condition is a geometric property under which the relative locations of the nodes inside a cell do not matter and the network is free of \textit{hidden nodes}. For the cases of saturated nodes and TCP-controlled long-file downloads, we provide accurate predictions of cell throughputs. Similar to Bonald et al (Sigmetrics, 2008), we model a multi-cell WLAN under short-file downloads as "a network of processor-sharing queues with state-dependent service rates." Whereas the state-dependent service rates proposed by Bonald et al are based only on the \textit{number} of contending neighbors, we employ state-dependent service rates that incorporate the the impact of the overall \textit{topology} of the network. We propose an \textit{effective service rate approximation} technique and obtain good approximations for the \textit{mean flow transfer delay} in each cell. For TCP-controlled downloads where the APs transmit a large fraction of time, we show that the throughputs predicted under the PBD condition are very good approximations in two important scenarios where hidden nodes are indeed present and the PBD condition does not strictly hold.