Zhang Yelei

1
Handling RFID Data

• Zhang Yelei
• 20 November, 2009

2
Presentation Outline

• RFID Introduction
• Data Processing
• Data Integration
• Questions

3
Presentation Outline

• RFID Introduction
• What is RFID
• Standards
• System architecture
• Applications
• Challenges
• Data Processing
• Data Integration
• Questions

4
What is RFID (1)

• RFID Radio Frequency Identification
• A new method/technology of remotely storing and
  retrieving data
• Need to use RFID readers and tags
• A little historical information
• First applied in World War II by UK
• First paper exploring RFID Communication by
  Means of Reflected Power by Harry Stockman
  published in 1948
• Not until 1990s it was widely deployed

5
What is RFID (2)
Roy Want, RFID a Key to Automate Everything
6
What is RFID (3)

• RFID Tags (transponder)
• Low cost, not application-specific
• Operate on frequencies ranging from 100KHz to
  beyond 2.5 GHz
• Type passive, semi-passive, active
• Majority are write-once/read-only, others offer
  r/w capability
• Readability influenced by factors like frequency,
  environment, tag position, antenna direction etc.
  
• Christian Floerkemeier and Matthias Lampes
  cards playing experiment illustrates some sources
  of errors in tag reading.

7
What is RFID (4)

• RFID Readers
• Portable or fixed
• Use serial port (RS232) or network
  interface/protocol (wired/ wireless connection)
  to communicate with computers
• Radio could be software defined

8
Related Standards

• Standards about frequencies and communication
• Identification cards and related areas ISO/IEC
  10536, ISO/IEC 14443
• Automatic identification and data capture
  technologies ISO/IEC 15961, ISO/IEC 15962
• Conformance ISO/IEC 18046, ISO/IEC 18047
• ETSI, ERO
• Standards about the data format on tags
• EPCGlobal focuses on the standardization of the
  data format
• EPC, electronic product code (64,96,256 bits
  long), is now the internationally accepted
  item-level code.

9
System Architecture
Savant mapping low-level data stream from
readers to a more manageable form, cleaning data,
supporting simple queries and installed standard
queries Central IS provide high-level services
that are easier for application to use.
10
Applications (1)

• Business applications
• Transport and logistics
• Supply chain management
• Agriculture
• Government applications
• Defense and security
• Library systems
• Consumer applications
• Personal welfare and safety
• Sports and leisure
• Shopping and dining out
• Smart homes

11
Applications (2)

• EPCGlobal Network
• A method of using RFID to share information in
  the global supply chain
• 5 components
• EPC (Electronic Product Code), ID system, EPC
  middleware, Discovery services, EPC information
  services (EPC IS)

12
Applications (3)
Source Sun and RFID
13
Challenges

• Reducing tag costs
• Global standards
• Frequency of tags and readers
• Other specifications
• IT infrastructure
• Data processing handling large amount of stream
  data online, effecient use of storage, network
  bandwidth, and so on.
• Integration with databases, data warehouses and
  enterprise applications
• Security issues

14
Presentation Outline

• RFID Introduction
• Data Processing
• Challenges of Handling Data Stream
• DSMS vs DBMS
• Projects
• Problems
• Solutions
• Data Integration
• Questions

15
Challenges of Handling Data Stream

• What is the data stream
• A potentially unbounded sequence of tuples
  (transactional data stream and measurement data
  stream)
• Data is continuous, infinite
• Most operations should be done online without
  interrupting data stream.
• Data recovery could also be a serious problem
• Computational resources are limited
• Real-time data stream requires efficient data
  handling
• Complex queries need to be performed nearly
  real-time

ATT Labs-research, Data Stream Query Processing
16
DSMS vs. DBMS

• DSMS
• DAHP model
• Deals with tuple sequences
• Complex queries executed real-time and online
• Database updated frequently
• Query persistent, plan adaptive, answer
  approximate
• DBMS
• HADP model
• Deals with tuple sets
• Complex queries usually executed offline
• Database relatively stable
• Query transient, plan fixed, answer exact

17
Research Projects

• Aurora (supports cq, ad-hoc query, and
  materialized view)
• Aims to better support monitoring applications
• Borealis (distributed SPE, QoS based techniques)
• A distributed stream processing engine based on
  Aurora and Medusa
• TelegraphCQ (focuses on hybrid, ad-hoc query)
• Intends to handle large streams of continuous
  queries over high-volume, hightly-variable data
  streams
• PSoup (focuses on both ad-hoc and continuous
  query)
• A query processor that supports both streaming
  data and streaming query
• STREAM
• A general purpose DSMS prototype
• GigaScope, Hancock, Nile, TinyDB, COUGAR

18
Problems

• Data models
• Window operations
• Query languages
• Query processing
• System optimization

19
Solutions (1) Data Models

• Relation-based Models
• Aurora stream type ( TS, A1,, An)
• PSoup
• STREAM a stream S is an unbounded bag of pairs
  lts,?gt, a relation R is a time-varying bag of
  tuples
• Object-based Models
• COUGAR and Tribeca data types are associated
  with methods

20
Solutions (2) Window Operations

• Why?
• Time/ordering is a very important aspect of
  streaming data
• Data processing is still based on a finite data
  set.
• How to define?
• Window can be time-based or tuple-based, or
  partitioned sliding window.
• Types
• Fixed, snapshot
• Landmark
• Sliding

21
Solutions (3) Query Languages

• Relation-based Languages
• CQL used by STREAM select from S1 Rows 1000,
  S2 Range 2 Minutes where S1.AS2.A and S1.Agt10
• Object-based Languages
• COUGAR select R.s.getTemperature() from R where
  R.floor3 and every(60)
• Others (Procedural Languages ??)
• Aurora 7 new operators like map, resample
  are defined

Golab and Ozsu, Data Stream Management Issues
----- A Survey
22
Solutions (4) Query Processing

• Use connection points to cach streaming data
  (Aurora)
• TelegraphCQ use OSCAR for the trade-off of
  quality and size of the data (from the disk)
• Attach data queues with operators
• In Aurora, queue is managed by successors
  pointers.
• Shared modules among different queries
• In STREAM, synopses is replaced by stub and store
  to reduce redundancy.

23
Solutions (5) System Optimization

• Data gathering
• Run-time statistics are gathered
• Inserting, combining, reordering operators
• Train scheduling, superbox scheduling (like batch
  operation)
• Load shedding
• Static analysis and delay-based dynamic analysis
  for overload detection
• By dropping tuples, or by value-based tuple
  filtering

24
Presentation Outline

• RFID Introduction
• Data Processing
• Data Integration
• Research gap
• Design considerations
• Questions

25
Research Gap

• Academic research
• Focuses on issues like processing ability,
  efficient deployment, antenna design, and so on.
• Lack of the emphasis on the effective interaction
  with data warehouses and high-level applications.
• Enterprise IS and data warehouse
• Emerged in 1980s, intend to deal with discrete,
  aggregated data, not continuous, real-time,
  single-item data.

26
Design considerations

• Manage data storage
• Which data should be saved and where
• Eliminate redundant data
• Handle historical data
• Query data
• Study business scenarios
• Identify typical on-site queries and data
  warehouse queries
• Real-time processing
• Design of triggers
• Link real-time events with business processes.
  (for example, BPEL process and web service)

27
Presentation Outline

• RFID Introduction
• Data Processing
• Data Integration
• Questions