IrisNet: An Architecture for a Worldwide Sensor Web

Gibbons, Karp, Ke, Nath, Seshan

sensor networking applications www internet service composition

@article{gibbons:ieee-pervasive2003,
  title={{IrisNet}: An Architecture for a Worldwide Sensor Web},
  author={Phillip B. Gibbons and Brad Karp and Yan Ke and
          Suman Nath and Srinivasan Seshan},
  journal={{IEEE} Pervasive Computing},
  year={2003},
  month={October--December},
  volume={2},
  number={4}
}

Working on architectures for constructing worldwide sensor webs

• Many, many sensors connected & accessible from the Internet
• Sensors here includes entire sensor networks, cameras, workstations, etc
  • Not just typical sensor motes, but a wide mix of devices

Problems here are not so much impoverished resources on the nodes but scalability, discovery, robustness, ease of use

• Trying to support rich queries over available data, composition, etc

Locality is important

• Most users of most data will be located relatively nearby the sources
• Data should therefore be stored near those sources

Information assurance is a large issue

• Policies defined by service users, service authors, those being sensed, those operating the sensor web

Data from sensors is published as XML

• XPath is used as the query language
• Data is stored as XML documents on Organizing Agents, which then apply received queries
• Any missing data, i.e. any not found in the XML tree, is fetched from descendent organizing nodes or sensors.
  • These are determined via the XML path, which specifies the relationships between the organizing agents

Databases ond organizing agents are also replicated to promote redundancy

• Cached query results are used for both performance & fault tolerance; queries may stipulate timeliness criteria

Describes several prototype applications

• Parking space finder: Monitors parking lots, maintaining database of open parking; users may query for spaces, receive directions
• Network monitoring: PlanetLab demo
• Coastal imaging: Variety of composition and filtering services work to fuse images from live cameras on Oregon coastline and detect visible features such as sandbars and riptides, data about which is archived for querying along with picture