We present an architectural framework, Garnet, which a data stream centric abstraction to encourage the manipulation and exploitation of data generated in sensor networks. By providing middleware services to allow mutually-unaware applications to manipulate sensor behaviour, a scalable, extensible platform is provided. We focus on sensor networks with transmit and receive capabilities as this combination poses greater challenges for managing and distributing sensed data. Our approach allows simple and sophisticated sensors to coexist, and allows data consumers to be mutually unaware of each other. This also promotes the use of middleware services to mediate among consumers with potentially conflicting demands for shared data. Garnet has been implemented in Java, and we report on our progress to date and outline some likely scenarios where the use of our distributed architecture and accompanying middleware support enhances the task of sharing data in sensor network environments.
Lyndell St. Ville, Peter Dickman