The design and evaluation of a query processing architecture for sensor networks

With the advent of small, battery-powered, wireless computing and sensing technology, it is now possible to monitor and observe the world at unprecedented levels of granularity. Networks of such devices typically consist of tens or hundreds of small, power constrained nodes deployed in remote locations which they are expected to monitor for months or years at a time. Such networks present significant new opportunities for the data management community. In this dissertation, we summarize the issues and opportunities associated with collecting and processing information from these wireless sensor networks, focusing on the performance and ease-of-use advantages of a declarative, query-based approach. We present the architecture of a query processing system and design of a query language specifically tailored towards power-efficient acquisition and in-network processing of sensor data, with a focus on issues related to minimizing unnecessary communication and sensing. We present many of these ideas in the context of TinyDB, a prototype database system which runs on networks of Berkeley Motes, and evaluate them using both performance micro-benchmarks and measurements from real-world deployments.