Transcript ppt file - IPFW.edu

ECET 581
Wireless Sensor Networks
Sensor Network Data Management
December 5, 2006
Fall 2006
http://www.etcs.ipfw.edu/~lin
References:
**Ch 6 Sensor Network Databases of the book: "Wireless Sensor Networks
- An Information Processing Approach," by Feng Zhao and Leonidas
Guibas, from Morgan Kaufmann/Elsevier inc, 2004
** Ch 10 Data Management of the book: Principles of Embedded
Networked System Design, by Gregory and William Kaiser, from
Cambridge University Press
**TMote Sky Web site: www.moteiv.com
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
1
Sensor Networks Data
Management


Traditional Database Management
System
Sensor Data & Database
• Sensor data organization
• Distributed storage
• Query & access



Wireless Sensor Database Design
Data Flow in Sensor Networks
High-level Reasoning
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
2
Traditional Database Management
System – an overview

Layered Architecture
• Database
• DBMS (Database Management System)
• Communication Network
• Client
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
3
Traditional Database Management
System – an overview


Database Schema – the structure and
constraints of the data format defined
through Data Definition Language (DDL)
Table
• Rows - record tuples
• Column – data attributes

DDL Compiler
• Translates these definitions into Metadata
• Metadata – a data structure describing the
structure of database and constraints
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
4
Traditional Database Management
System – an overview

Transactions - a sequence of operations
• Retrieving data
• Database updating
• Producing new tables

ACID Properties
• Atomicity – all of the transaction takes place or
nothing changes in the DB
• Consistency – correctness (from begin to end)
• Isolation – the transaction is concealed from
other transactions until it commits
• Durability – update effected by a transaction
survive until affected by other transactions
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
5
Traditional Database Management
System – an overview

Query
• Query Language – SQL (Structured Query
Language)

Join, group-by, sort, scan
• Query processor
• Execution engine

Data storage
• Geographically distributed at various locations
• Robust to failure
• Web-based access, remote access
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
6
Sensor Data & Database

Data Info of Sensor Network
• Time-stamped measurements of physical
phenomena

Heat, sound, light, pressure, motion, etc
• Abstract data representation

Detection, classification, tracking, etc
• Sensor Description: location, types, etc

Database?
• Traditional DBMS system
• DB within the network
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
7
Sensor Data & Database

Challenges
• The system is highly volatile
• Relational data tables are not static – new data is
continuously being sensed
• High energy cost in communication – encourage in
network query
• Long delay, need to continuously monitor the rates and
availability of data
• Maintain high-level statistical summaries: Limited
storage on node and high communication cost
• Sensor queries: queries from others, and self generated
queries
• Metrics of performance
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
8
Distributed Database – Wireless
Sensor Network





A set of database sites, each is a complete
database system
Local DB – store frequently used data to
minimize long-distance retrieval of large
quantities of data
Fault tolerant
Primary copy, Secondary copies
Data object naming:
•
•
•
•
Creator ID
Creator Site ID
Local name
Birth site ID
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
9
Data Flow in Sensor Networks

Data Management
• Queries should direct


collection, storage, and transport of data and
geographical and temporal scope
Signal processing, aggregation, and routing
strategies
• Must still consider




Concurrency
Integrity
Security
Efficiency
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
10
Data Flow in Sensor Networks






Dealing with the infinite
Probabilistic data collection
Sensor networks as data
compression engine
Interest diffusion
Adaptive data collection
Standards vs. custom design
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
11
TinyDB Features



Metadata management
High-level queries: SQL like
Network topology
• Tracking neighbors, and maintaining routing
tables to determine how to deliver data
efficiently

Multiple queries
• diff sample rate and access diff sensors

Incremental deployment using query
sharing
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
12
High-Level Reasoning




Data Mining
Logic: if, then, else, not, therefore
Bayesian Networks – Belief networks
Language
• Prolog
12/5/2006 - updated
WSN Mote- MCU & Sensor Hardware
13