RealTime Sensor Data Processing on Grid Environments

1
Real-Time Sensor Data Processing on Grid
Environments
Eui-Nam John Huh Kyung Hee University johnhuh_at_khu.
ac.kr
2
Contents

• Why Sensors on Grid?
• Sensor Network Challenging Problems
• Data Archive/Filter
• Dynamic Sensor Topology
• Dynamic Real-Time Computing
• Sensor Grid Requirement
• Sensor Grid Simulator
• Mobile Gateway
• Discussion

3
Why Sensor Grid?

• Current Grid Architecture Community efforts to
  establish well specified Grid capability and
  components (behaviors).
• Advanced IT technologies require an efficient
  solution to the real-time (on-demand) problem
  using supercomputers or clusters that gathers
  tera-byte data from real-environments by various
  sensors and needs to process them in real-time.
• That is, Grid technology will be a strong
  infrastructure for those problems as well as
  science and business problems.
• This research group specifies service
  requirements for the Sensor Grid in order to
  support OGSA specification to be the general
  purpose grid.

4
Grid Today
Quick Review

• Grid (Bag of Technology) includes most of
  Computer Techs.
• Parallel/Distributed processing
• High Performance Network
• Security
• Super Computing Platform/Clustering Tech.
• Current Grid works in main for
• Scientific Simulation (1997-2002)
• Business Solution (2003-2005)

5
Grid Future
Quick Review

• How to Include New Technology
• Wireless Sensors (Network) Technology
• Ubiquitous environment
• Real-Time Data Stream from diverse sensors
  located in real environments
• Clustered Small Hand-held devices w/ camera or
  sensors (RFID)
• All around has full of sensors such as badge
  readers, credit card readers, doctors instrument
  (pulse, blood pressure) on all places
• Interactive response is still required
• Grid Technology will handle these issues as IT
  Infrastructure ?
• If yes, what components need to be updated
  or inserted?

6
(No Transcript)
7
Conceptual View of Grid Infrastructure
Quick Review
8
Moving Towards Business Real Time
9
Usage of Sensor Based Computing

• Asset Management (RFID basic environmental
  data)
• Supply chain
• Manufacturing and construction
• Luggage handling
• DOD (Department of Defense)
• NASA dangerous chemicals
• Emerging applications building Grid communities
• (Environmental) data ORNL (SensorNet)
• Healthcare services UK healthcare

10
RFID Physical World ? IT World
Sensor Data Archive
RFID Tag
Reader
Hardware Physics
Information Technology
Antenna
Chip
Edge Server
Applications
11
Dynamic Real-Time Computing
Deterministic Stochastic Dynamic
12
Dynamic Real-Time Computing

• Dynamic Real-Time Computing
• Unknown worst-case scenarios
• Large variances in sizes of data and event sets
  (hence large variances in resource requirements)
  in run-time
• Unbounded resource requirements
• Various arrival rates
• Cannot be characterized by time-invariant
  statistical distribution
• e.g. mission critical system, air traffic
  control system,Telematics, and robotics
• gt React to the environments in a timely manner

13
Towards The Real Time Enterprise
BACKEND SYSTEMS
Sensors Actuators
ACCESS
Business Intelligence
Portal
Alerts
Packaged Apps

• ANALYZE

Packaged Apps
Business Processes
Business Activity Monitoring
Integration
MANAGE
CAPTURE
Legacy
Sensor Data Archive
Enterprise Manager
Security
Edge Server
Other
14
The World of Sensor Based Computing

• Increases significantly the fidelity of the
  representation of the physical world
• Captures sensor readings in time
• Associates temporal topologies to sensors
• Static layout of a warehouse
• Dynamic interaction of measurements
• Creates temporal and spatial description of the
  physical world
• Increases significantly the ability to react to a
  changing world

15
Edge Processing Challenges
EVENT SOURCES

• Many types of event sources and sensors
• RFID (handheld, portal, truck mount, etc.),
  barcode, location, moisture, temperature,
• Many RFID and sensor vendors
• Alien, Intermec, Matrics, ThingMagic, SAMsys,
  Zebra
• Many different frequencies, protocols, interfaces
• LF, HF, VHF, UHF, Microwave
• Ethernet, PCMCIA, RS-232
• Many devices in different locations to manage
• Device health, device upgrade
• Central management
• Information Capture
• Data format not consistent
• Data is granular and redundant
• Information based on observations from multiple
  sensor sources

16
Edge Server Functions
EVENT SOURCES

• Filter Framework
• Remove unwanted or low-level events
• Extensible filter architecture
• Generates high level events and optimize flow and
  responses
• Filters
• PassThru Filter for portal readers or choke
  or point reads
• Shelf Filters for shelf and proximity readers
  producing enter/exit events, or period read
• PalletPass, PalletShelf Aggregation filtering
  for both Pass and Pallet filters
• CrossReader All filters can be applied to
  groups and individual devices
• CheckTag Test tag to verify reader health
• Device Groups
• Allows administrators to logically group devices
  together
• Group devices for logical processing and filtering

17
Towards The Real Time Enterprise
BACKEND SYSTEMS
EVENT SOURCES Actuators
ACCESS
Oracle
Business Intelligence
Portal
Alerts
E-Business Suite

• ANALYZE

Packaged Apps
Business Processes
Business Activity Monitoring
Integration
MANAGE
CAPTURE
Legacy
Enterprise Manager
Edge Server
Security
Other
Sensor Data Archive
18
Sensor Data Archive
External Data
Organize
Sensor Reading
Process/ Analyze
Sensors Topologies
Disseminate
19
Sensor Data Archive Sensor Readings

• Sensor readings
• Temporal representation of sensor measurements
• Point-in-time
• Period
• Raw data temperature, location, .
• Derived data RFID tags (number, speed,
  direction, )
• These are observations not events
• Events are interpretations of observations
• Performance, scalability, security, are
  essential

20
Sensor Data Archive External Data

• Temporal information from non (local) sensors or
  applications
• Data received from external sensors
• Data received from applications and/or business
  partners
• Containment hierarchies content of pallets,
  boxes,
• Protected data unpaid items
• All data can be received Ad hoc, scheduled, or
  event driven

21
Characteristics of Sensor Network

• Data stream from multiple sensors with various
  constraints
• Periodic or Aperiodic data arrivals
• Classification of sensors
• Locating sensors
• Tracking the trajectory of sensors
• Predicting the location/trajectory of sensors
• Putting confidences on the accuracy of tasks

22
Sensor Grid

• Tera-Bytes Data Stream On-Line Processing
• Multi-Dimensional Requirements (Combined
  Constraints correlating RFID and Temperature
  (car (refrigerator (item))-gtbuilding-gtroom-gtrefrig
  erator (item))
• Dynamic Sensor Network Topology Tracking
• Activation of Control Grid

23
Sensor Grid Today
- Dedicated to C2 node
24
Sensor Grid Future
- Interoperability
25
OGSA Consideration for Sensor Grid

• Connection and communication (common control
  protocol)
• - multiple active sensors
• Locating / Discovering Sensors
• Topology is changed dynamically
• Time / Space Constraints
• changes by Real-time
• life time issues
• separation in space and time
• Resource Management
• Real-Time Wrapping
• Predictability
• Dynamic Workload Handling Model
• Scale-up is not enough operation

26
Grid Simulator Model

• Intelligent agent base
• Real-time sensor network simulator
• User interface
• each sensor component can be controlled by
  resource manager or user
• Dynamic scalable application model
• Specification language for different sensors and
  requirements including
• Constraints security level, time, QoS
• Characteristics of each system
• Requirements throughput, performance

27
Sensor Grid Simulator
Scenario file
Doctrine for event
Display
Sensor
Analysis Clusters
Filter
Filter Clusters
Evaluate Decide

events
Data Archive
Sensor
Action Data Clusters
Actuator (Shooter)
Disseminator


Sensor
Actuator (Shooter)
Action (Control)
28
Sensor Grid Gateway
29
Sensor Grid Gateway
30
Discussion

• QA
• Further questions to johnhuh_at_khu.ac.kr