In its evolution to provide ever higher data rates, the Wi-Fi standard has incorporated sophisticated PHY-layer techniques, which has in turn increased the complexity of network-wide interference relationships. Proper modelling of the resulting inter-device interactions is crucial for accurately estimating Wi-Fi network performance, especially in the contemporary context of traffic and network densification. Event-driven simulators like the open-source ns-3 are in principle able to capture these interactions, however it is imperative to validate, against experimental results, whether their underlying models reflect the network behaviour in practice. In this paper we first perform experiments in a large-scale indoor testbed to validate the IEEE 802.11ac Wi-Fi model in ns-3, for various channel width and allocation configurations. Our results show that ns-3 captures Wi-Fi co-channel interactions with reasonable precision, but fails to model adjacent channel interference (ACI), which our experiments show to be critical in dense networks. We therefore propose and implement an ACI model in ns-3. Importantly, our model successfully captures the qualitative behaviour of the CSMA/CA mechanism when transmissions on adjacent channels occur. Further, our ACI implementation significantly improves the accuracy of both the network and per-device throughput estimates for the considered dense IEEE 802.11ac network compared to the basic ns-3 Wi-Fi model without ACI. For example, without ACI modelling, ns-3 overestimates the aggregate network throughput by up to 230%, whereas with our ACI implementation the aggregate throughput estimate is no more than 65% higher than the experimental results.
