Models and systems for understanding wireless interference

With the prolific growth in the usage and the number of wireless devices, the problem of designing high performance and reliable wireless networks is of great importance today. A crucial step forward in designing such networks is to improve our understanding of wireless interference — a phenomenon that remains the primary source of performance bottlenecks and unpredictability in today's wireless networks. To this end, this dissertation makes contributions in developing systems and models that help us better understand interference in wireless networks. We start by studying wireless losses. Wireless losses can be due to weak signal at the receiver, or interference from other wireless devices such as WiFi devices or non-WiFi RF devices. To isolate the losses emanating from interference due to WiFi devices, we design COLLIE, a system that helps distinguish between losses due to weak signal and those due to WiFi interference. We then focus our attention to the problem of non-WiFi interference. Since WiFi cards do not have the ability to decode a non-WiFi device's transmission, we designed \as∼a software system that runs on top of a mainstream wireless card and helps detect different non-WiFi devices. We then designed \fusion, a system that helps detect the exact source of interference, quantify the impact of such interference, and also help physically pin point the non-WiFi interferer's location. All these systems are based on post-mortem analysis of wireless losses. In the last part of this thesis, we also explore an alternative method of modeling wireless interference in the context of links using flexible channels. These models are predictive in nature and help us understand the impact of wireless interference before the event of a packet loss. We believe that the models and systems presented in this thesis are useful in understanding wireless interference in today's networks as our solutions can be natively implemented in present-day WiFi hardware. We hope that our contributions would serve as useful building blocks in designing more sophisticated tools to better understand wireless interference and pave way for building future wireless networks that are more reliable, predictable and manageable.