The acquisitional query processing system for sensor networks described in this paper is a new development in the field of acquisitional query languages. Running on the Berkeley “mote” platform, on top of an operating system called TinyOS, TinyDB is a query processor designed for sensor networks that has control over “where, when and how often data is physically acquired.”

The discussion starts with a presentation of the sensor devices and their power consumption, the operating system running on these sensor devices (TinyOS), and the communication inside sensor networks. On these sensor devices, on top of TinyOS, there is an acquisitional query language running, called TinyDB, that collects data in the form of sensor tuples. TinyDB uses basic structured query language (SQL) with extensions that deal with the particularities of the sensor devices, mainly the sample period (the time between the moment when the nodes initiate the data collection and transmission in a real-time fashion, and when they stop the transmission).

Other extensions to the classic SQL used inside TinyDB concern the aggregation queries, and also event-based queries (where the events are those mechanisms that initiate data collection). The lifetime-based queries will be useful in environmental monitoring scenarios, where scientists are concerned with the duration of the network. The model of TinyDB uses a cost-based optimizer to choose a query plan that will require the lowest degree of power consumption. This is accomplished by the use of a catalog of metadata describing the node’s local attributes, events, and user-defined functions.

Next, the paper describes the dissemination of the query into the network. The query begins with a broadcast from the root of the network. Each node listens to the query, and must decide if it applies locally, or needs to be further broadcast to its children. This decision is based on a new data structure proposed by the authors: a semantic routing tree. This routing tree allows each node to efficiently determine if its children need to participate in a given query.

Finally, the authors present details on query execution, prioritizing data, and adapting sampling and delivery rates. Using techniques like snooping, TinyDB can still transmit the most relevant results when the power or bandwidth is limited.

Although the paper is very interesting, and describes the newest techniques in data acquisition and processing, its audience is quite limited; it is mainly intended for those scientists and researchers who implement and supervise battery-powered sensing devices.