IrisNet: An Architecture for Internet-scale Sensing Services Suman Nath, Amol Deshpande, Yan Ke, Phillip B. Gibbons, Brad Karp, and Srinivasan Seshan   View Paper (PDF)   Return to Advanced Models and Languages/Architectures for Data Analysis (Session D3) Abstract We demonstrate the design and an early proto- type of IrisNet (Internet-scale Resource-Intensive Sensor Network services), a common, scalable networked infrastructure for deploying wide area sensing services. IrisNet is a potentially global network of smart sensing nodes, with webcams or other monitoring devices, and organizing nodes that provide the means to query recent and his- torical sensor-based data. IrisNet exploits the fact that high-volume sensor feeds are typically attached to devices with significant computing power and storage, and running a standard op- erating system. It uses aggressive filtering, smart query routing, and semantic caching to dramat- ically reduce network bandwidth utilization and improve query response times, as we demonstrate. Our demo will present two services built on Iris- Net, from two very different application domains. The first one, a parking space finder, utilizes webcams that monitor parking spaces to answer queries such as the availability of parking spaces near a user's destination. The second one, a distributed infrastructure monitor, uses measure- ment tools installed in individual nodes of a large distributed infrastructure to answer queries such as average network bandwidth usage of a set of nodes. ©2004 Association for Computing Machinery