Inference of Wi-Fi busy time fraction based on Markov chains

IEEE 802.11 has emerged as a vital wireless network access technology for mobile devices. By providing the potential for high connectivity speeds, this technology has led to a huge rise in the number of access points (APs). In such environments, mobile devices have the choice to join several Wi-Fi networks. Despite its importance to user Quality of Experience (QoE), the AP selection is still trivial since it focuses at best on the received signal strength if not only the user's history. Crucial metrics that capture the overall dynamics of the AP load condition, such as the network load, are not taken into account. In this paper, we propose to use the Busy Time Fraction (BTF) as a metric to choose the best AP to attach to. The BTF level of a given channel is inferred based on the frame aggregation scheme proposed since the 802.11n standard. In this regard, we build a proof of concept system, FAM (Frame Aggregation based method) that leverages the theoretical frame aggregation levels of a probe traffic returned by two analytical Markovian models and the measured ones in order to estimate not only the BTF but also the nature of the traffic. We validate the accuracy of our proposed approach against ns-3 simulations under several scenarios.