Traffic Sensing and Characterization in Multi-channel Wireless Networks for Cognitive Networking

Traffic sensing and characterization is an important building block of cognitive networking
systems, however, it is very challenging to perform traffic characterization in multichannel multi-radio wireless networks. Due to the presence of network traffic in multiple
channels, the existing count-based packet sampling methods demand continuous capture
on each channel to be effective; this requires a dedicated wireless interface per channel,
and hence the existing sampling methods require a very expensive infrastructure and have
poor scalability. Time-based sampling methods, on the other hand, offer a cost-effective
and scalable solution by reducing the amount and cost of the resources necessary to monitor and characterize the wireless spectrum.
The contributions of this paper include the following: (i) a discussion of packet sampling
techniques for traffic sensing in multi-channel wireless networks, (ii) a comparison of various time-based sampling strategies using the Kullback–Leibler divergence (KLD) measure,
(iii) a study on the effect of the sampling parameters on the accuracy of the sampling strategies, (iv) development of sampling accuracy graphs for easing the process of best sampling scheme selection in multi-channel wireless networks, (v) the proposal of a new
metric (traffic intensity) which estimates the busyness of channels by taking into consideration not only the successfully received packets but also corrupt or broken packets,
(vi) implementation of time-based sampling in a prototype traffic sensor device for
multi-channel traffic sensing in IEEE 802.11 b/g networks, and (vii) characterization of a
campus IEEE 802.11 network environment in a spatio-temporal–spectral fashion using
sampled traffic traces collected by traffic sensors.