T Querying the Physical World

sensors, actuators, and mobile devices is transforming the physical world into a computing platform. We will soon find computing power, memory, and communication capabilities on temperature sensors and motion detectors, on door locks, light bulbs, and alarms, on every cellular phone, in every vehicle, and soon in every person's wallet or on their key ring. Emerging networking techniques ensure that devices are interconnected and accessible from local-or wide-area networks [1]. Using this new computing platform, users interact with portions of the physical world. In a large class of applications, users monitor phenomena in a given environment. Examples of monitoring applications include gathering information in a disaster area, supervising items in a factory warehouse, or controlling vehicle traffic in a large city [2, 3]. Let us take the concrete example of an existing flood detection system. For about twenty years now, the ALERT system has been deployed in several US states (http://www.alertsystems.org). A typical ALERT installation consists of several types of sensors in the field: rainfall sensors , water level sensors, weather sensors, etc. A predefined set of data is regularly extracted from each sensor, transferred to a central site and stored in a database system. Users query the database system through a graphical user interface. Here are some example queries that users can express: " For each rainfall sensor, display the average level of rainfall for 1999, " Display the current level of rainfall for all sensors in Tomp-kins County, or " Every hour, display the location of the sensors where the level of rainfall is greater than 250 mm. " The example of the flood detection system emphasizes that monitoring is best described in a declarative manner: users submit queries concerning a device network regardless of its physical structure or its organization. In monitoring applications , users typically ask three kinds of queries: • Historical queries: These are typically aggregate queries over historical data obtained from the device network, e.g., " For each rainfall sensor, display the average level of rainfall for 1999. " • Snapshot queries: These queries concern the device network at a given point in time, e.g., " Retrieve the current rainfall level for all sensors in Tompkins County. " • Long-running queries: These queries concern the device network over a time interval, e.g., " For the next five hours, retrieve every 30 seconds the rainfall level for all sensors in Tompkins County. " The existing …