WhiteCell: Energy-Efficient Use of Unlicensed Frequency Bands for Cellular Offloading

Cellular networks have addressed the multi-fold increase in traffic demand through various approaches from increasingly smaller cells to offloading demand to unlicensed spectrum, such as WiFi and TV white spaces. The latter approach has a tremendous cost benefit as unlicensed hardware can be co-located with existing cellular infrastructure. However, in these situations where demand is the greatest, WiFi spectral activity could be high and the total number of available white space channels are often inversely proportional to the population due to the existence of TV broadcast channels. Moreover, the additional hardware induces a higher energy demand of the cell site. In this work, we perform extensive spectral measurements of unlicensed bands in a major metropolitan area and crowdsource user mobility to consider potential swings in network load. Using these measurements, we design a queuing-based approach to serve these network demand dynamics in an energy-efficient manner according to differing qualities of service. In particular, we perform in-field experimentation of the diurnal spectrum activity across white space (54-806 MHz) and WiFi (2.4 and 5.8 GHz) bands in typical settings across the Dallas-Fort Worth metroplex. We also consider mobility patterns from Android-based crowdsourced measurements in four major Texas cities to infer the change in traffic demand induced by users. Driven by both data sets, we propose a Greedy Server-side Replacement (GSR) algorithm to estimate the power consumption for the use of unlicensed bands in cellular offloading. In doing so, we find that networks with white space bands reduce the power consumption by up to 513% in sparse rural areas over WiFi-only solutions via measurement-driven numerical simulation. In more dense areas, we find power consumption reduction across a 24-hour period to be, on average, 24%, 44%, 63% over WiFi-only offloading with one to three white space channels, respectively. Finally, we consider the quality of service impact on power consumption and find that up to 151% of the power can be saved with only a slight relaxation of waiting times.