As mobile and wireless networks become the dominant user access technology, they face an explosion in demand for high-bandwidth Internet and cellular connectivity. Additionally, wireless, both cellular and broadband, has become the de facto access technology at the Internet's frontier in developing contexts and other resource-limited environments due to underlying economic factors. In order for wireless and mobile networks to maintain pace with demand, and achieve global connectivity, we must empirically study existing networks and use those findings to inform new system designs across the telecommunications hierarchy.
In this thesis, we measure and characterize real-world operational networks, both cellular and wireless broadband, so that we can later use insights gleaned from this analysis to inform system designs aimed at ameliorating the performance inadequacies we witness. Our work includes a unique glimpse into cellular infrastructure performance in an operational refugee camp from the user's perspective. Our analysis shows that the networks we study often provide users with diminished performance due to factors such as oversubscription. Given our insights in the refugee camp context as well as cellular traces collected in rural Guatemala, we design multiple systems for improving coverage by identifying failure of nearby commercial infrastructure and offloading user traffic during the presence of congestion onto a local cellular network.
Wireless Internet Service Providers (WISPs) represent a model for delivering connectivity to rural areas for relatively little cost. Unfortunately, WISP networks are challenging to design and are often created and maintained by individuals with little networking expertise, leading to brittle, tree-like topologies and poor performance. We design a prototype system for automated planning of WISP network topologies using topographic data and the preliminary design of a protocol that enables local multipath specifically tailored for the challenges unique to the WISP context.
Wireless spectrum is a scarce, finite resource that is a fundamental requirement for wireless communications. Recent advances in agile radio technologies have enabled the promise of dynamic spectrum access models, where unoccupied spectrum can be shared by primary and secondary users. In order to realize the promise of DSA, we must first have knowledge of the spectrum occupancy in a given location. With this goal in mind we design AirVIEW, a one-pass transmitter detection algorithm that is resistant to noise.