NeTS-NECO: Robust, Delay-Tolerant Sketches for Aggregating Sensor Data Streams
Emerging data-intensive mobile sensor networking applications promise a close observation of the world around us at a relatively low data acquisition cost. In these applications, it is a challenge to aggregate observation streams from distributed sources, due to the bandwidth and delay constraints of the network, and the energy costs of data transmission. This project develops techniques for processing voluminous sensor data streams within the resource constraints imposed by a distributed infrastructure-less network.
Techniques are designed to summarize the observation streams into small space "sketches". By disseminating sketches rather than the raw data, it is possible to trade-off reduced communication cost for approximate answers to user queries. The project investigates the technical challenge of making the sketches robust and suitable for transmission over unreliable links. The sketches are delay-tolerant, i.e. they can tolerate long and variable network latencies. Various classes of queries are supported including: basic aggregates of numerical and categorical data, multi-dimensional temporal and geographic aggregates, and time-decayed aggregates. The design of sketches is informed and inspired by the recent rapid progress made in the area of massive data stream processing. The project will result in new algorithms and data structures for sketching sensor data with provable guarantees on the quality of answers. The research is expected to benefit data-intensive sensor networking applications that are critical to our society, such as earthquake, pollution, and traffic monitoring.
