Online Query Processing and Data Analytics on Sensor Data with AnduIN
- Speaker: Prof Kai-Uwe Sattler (Technische Universitat Ilmenau)
- Host: Alvaro A. A. Fernandes
- 9th February 2011 at 14:15 in Lecture Theatre 1.4, Kilburn Building
Sensor networks have evolved to a powerful infrastructure component for event monitoring in many application scenarios. Among the possible data processing paradigms, online processing of sensor data streams is a challenging task due to the theoretical infinite nature of streams and the fast arrival rate of data prohibit storing and multiple reading of data. In addition to simple operations like filtering and aggregation, an important task in this context is data mining - the identification of relevant information and patterns. Applications of such analytical operations range from basic data cleaning tasks to complex pattern mining including clustering and text processing, e.g. for Twitter data streams. In this talk, we present our system AnduIN for developing, deploying, and running applications of sensor data processing. The system consists of a data stream processing engine, a library of operators for sensor-local processing, and a box-and-arrow editor for specifying data processing tasks including data mining. In particular, we discuss the query planning and adaptive execution strategies used in AnduIN, which employ a multi-dimensional cost model taking energy consumption into account.