Managing Data Streams Originating in Wireless Sensor Networks

1
Managing Data Streams Originating in Wireless
Sensor Networks

• Lyndell St. Ville
• Dept of Computing Science
• University of Glasgow
• 24th November 2004
• lsv_at_dcs.gla.ac.uk
• http//www.dcs.gla.ac.uk/lsv/garnet.html

2
Presentation overview

• Introduction to sensor networks
• Growth trends and impacts
• Providing shared access to sensors
• Introduction to the Garnet Architecture
• Upgrade path and interaction model
• Summary

3
What is a sensor network?

• A generic term used to describe a collection of
  monitoring devices which are deployed within an
  environment to detect phenomena of interest.
• Examples (We will see actual ones later)
• A constellation of spy satellites or space probes
• The computers in the machine room
• Water-level sensors which detect river flooding
• Buoys measuring the ocean currents

4
What makes them interesting?

• Devices deployed within the environment
• Monitor local phenomena
• Report observations to distant operator
• Opportunity to react to detected events
• Adjust sensor behaviour
• Take other action (drive an actuator)
• Opportunity to improve measurements by sensor
  collaboration

5
Useful features and properties

• Widespread and dense coverage
• Large numbers of devices
• Dispersed over large area
• Untethered operation of devices
• Wireless ?, mobile ?, small ?, and low-cost ?
• Very limited energy resources
• NB Resource constraints affect most other
  aspects of sensor networks

6
Other properties

• Boosted by advances in technology
• Miniaturization, ubiquitous computing
• Predetermined usage of sensors
• Bespoke applications (data consumers)
• Consumers may reside in separate, distant, wired
  network

7
Future trends

• Reducing cost of components
• Expansion of existing sensor networks
• New deployments of sensor networks
• Increasing presence of data sources
• Attracting attention of software developers
• Unanticipated uses of sensor data

8
Examples of sensor networks

• Habitat monitoring
• Zebranet
• wildlife tracking system
• Great Duck Island
• habitat monitoring
• Road traffic monitoring
• SmartTrek
• transit guide and congestion information system

9
Examples of sensor nodes

• mote -- University of California at Berkely

10
Example 1 ZebraNet Wildlife Tracking System

• Princeton University biologists tracking animals
  in natural habitat
• A power-aware wireless ad hoc sensor network
• No cellular wireless infrastructure
• Peer 2 Peer operation
• Data Aggregation within the network
• Intermittently mobile base station
• Land Rover, airplane

11
Example 2 Great Duck Island Habitat Monitoring
System

• Intel Research,and University of California at
  Berkeley
• A habitat monitoring system
• Facilitates non-intrusive, non-disruptive
  monitoring of sensitive wildlife and habitats
• Island is rigged with sensor devices
• Data routed through on-island gateway, via
  satellite to university and made available on the
  internet

12
Great Duck Island Networking Overview
13
Example 3 SmartTrek Travel information system

• Washing State - Department of Transportation
• Transit guide and congestion information system
• City buses outfitted with transponders
• Data used to track traffic conditions to assist
  commuters
• Applications
• Trafnet - real-time Seattle traffic conditions
• Busview - check status of Metro buses.

14
SmartTrek
15
More
16
Overview of examples

• What do they have in common?
• Owned and controlled by a single organisation
• Bespoke, single-purpose
• Data mainly flows out of the sensor network
• Reprogramming (retasking) occurs en-masse
• Homogeneous sensing devices.
• Data consumption pre-determined
• Limited interaction with sensor devices
  (actuation)

17
Implications for growth

• Unanticipated uses of the data
• Unexpected users and requirements
• Increasing complexity of requirements
• Increasingly heterogeneous sensor devices
• Not sensible to replace entire network
• Centralized ownership/control is difficult
• Increasing need to
• Share sensor data, interact with sensor devices

18
Factors inhibiting data sharing

• Data sensitivity and security
• Device access limitations and side-effects
• Access costs, processing overheads
• Capacity, energy constraints
• Overhead in managing devices
• Complexity increased by heterogeneity
• Mission critical concerns
• Conflict with main purpose
• Limited extra capacity

19
Garnet project goals

• Exploiting sensor systems
• Managing data streams
• Controlled interaction with sensor devices
• Other related goals
• Extending deployed systems to secondary users
• Access control on change requests to sensors
• Monitoring operation of consumer processes
• Facilitating separation of concerns of
  stakeholders
• To reduce the latency of data changes requested
  by consumer processes

20
Watercourse example
21
Garnet architecture - overview

22
Key benefits

• Encourages separation of concerns of stake
  holders
• Enhances simplicity
• Sensor devices manipulated via a data-stream
  abstraction
• Data consumers are mutually unaware
• Supports different levels of importance
• System-level, 3rd-party consumers
• Priority mechanisms to override standard
  operation in case of crisis

23
Generalised view of access to sensor networks

• Maintains security
• Limited access or exposure to underlying sensor
  devices
• Risk in providing direct access to sensors

24
Existing approaches

• Direct interaction with sensor nodes
• Virtual machines in each sensor node
• Good for security
• Node-level middleware
• Useful for reprogramming the nodes
• Database abstractions
• Node-level query proxy service
• Sensor proxies

25
A data-stream centric solution

• Upgrade based on Garnet architecture
• Application-level middleware
• Consumers provided with opportunity to interact
  with sensor stations
• Requests for changes managed by middleware
• Overrides still permitted during crisis conditions

26
Upgrade overview

• Data received and routed to Garnet

27
Upgrade detailed

• Change requests processed by Garnet

28
Upgrade summary

• Consumers permitted independent access to sensor
  devices
• Can receive data updates
• Consumer updates and change request may be
  preempted by higher-priority requests
• Network owner independence

29
Summary

• Data stream abstraction
• Middleware approach to sharing and controlling
  sensor data
• Extensible platform

30
The End
Managing Data Streams Originating in Wireless
Sensor Networks
Lyndell St. Ville lsv_at_dcs.gla.ac.uk http//www.dc
s.gla.ac.uk/lsv/garnet.html