UNIVERSITY OF NIVERSITY OF MASSACHUSETTS ASSACHUSETTS, A , AMHERST • MHERST • Department of Computer Science Department of Computer Science SPIRE: Scalable Processing of RFID Event Streams Yanlei Diao University of Massachusetts, Amherst Joint work with Richard Cocci and Prashant Shenoy
16
Embed
U NIVERSITY OF M ASSACHUSETTS, A MHERST Department of Computer Science SPIRE: Scalable Processing of RFID Event Streams Yanlei Diao University of Massachusetts,
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science
SPIRE: Scalable Processing of RFID
Event Streams
Yanlei Diao
University of Massachusetts, Amherst
Joint work with Richard Cocci and Prashant Shenoy
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 2
UMass RFID Research Center
Kevin FuRFID security & privacy
Yanlei DiaoRFID data management
Wayne BurlesonSecure RFID Hardware
Mark CornerRFID locationing and
mobile readers
Prashant ShenoyRFID software systems
UMass Center for Advanced RFID Research
http://rfid.cs.umass.edu
5 faculty + 9 students
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 3
Research Thrusts
Embedded devices & hardware thrust
Data management middleware thrust
Application thrust
Secu
rity
&
Pri
vacy
Rob
ust
ness
&
Sca
lab
ility
Contact-less payments
Contact-less payments
Precise locationing at home/office
Precise locationing at home/office
Healthcare processmanagement
Healthcare processmanagement
Trace-&-trace in supply chain
Trace-&-trace in supply chain
Monitoring &anomaly detection
Monitoring &anomaly detection
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 4
Real-time Visibility using Event ProcessingReal-time Visibility using Event Processing
Monitor, Alert, Correct, Control, Improve
Data management middleware thrust
Data Management Middleware
Tracking Individual Objects on a Large ScaleTracking Individual Objects on a Large Scale
Track-and-Trace
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 5
Challenges
Data-information mismatch RFID data <tag_id, reader_id, time> Meaningful, actionable information
Incomplete, insufficient, misleading data Missing data Overlapping read ranges Location unclear Containment unclear
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 6
Challenges
Scalability Large-scale deployment:
Dozens of locations Thousands of readers Millions of objects
Unprecedented volume of data Low-latency
Anomaly detection: up-to-the-second information
Online data warehousing: frequent updates
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 7
Historical Data
4. Event DBTrack & trace
Overview of SPIRE
RFID DevicesRFID Devices
Continuous queries Results SQL queries Results
Data/Queries
1. Data Cleaning1. Data Cleaning Incomplete data, misleading data
Live Data
3. Complex Event Processor
3. Complex Event Processor
Monitoring, anomaly detection
2. Data Interpretation, Compression2. Data Interpretation, Compression Insufficient data, redundant data
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 8
Data Cleaning
Raw RFID Stream
Anomaly Filtering
Temporal Smoothing
Time Conversion
Deduplication
Removes bad data
Handles missing data
Adds logical timestamp
Resolves misleading
data
For details see [CocciDS07, GyllstromWC+07, JefferyGM06].
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 9
Location-Containment View
A physical world Objects, locations, object containment at
time t RFID readings
Isolated, fragmented views of the physical world
An integrated location-containment view View evolves as new readings arrive View is used to interpret (give meanings to)
the readings
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 10
View Maintenance
View(time T, location A)
1, 6, 2, 4, 51, 6, 2, 4, 5
+ Stream(T+1, A)
7, 3, 87, 3, 8
+ Stream(T+2, B)
View(T+2, B)
?X
View(T+1, A)
?
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 11
Data Interpretation & Compression
Interpretation of low-quality, insufficient data Location unclear Containment unclear
Data Compression- Location compression- Containment compression- Compression vs. real-time anomaly detection
Archival to event database for track-and-trace Querying over both history and current state
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 12
Complex Event Processing
Real-time translation from data to actionable information Filtering Correlation Aggregation Transformation Predication
Proactive, adaptive systems Monitor, alert, correct, control,
improve
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 13
SASE Event Processor
Computation Complexity
Misplaced InventoryEVENT SEQ(Shelf_Reading x,
Shelf_Reading y, !(ANY(Register_Reading, Shelf_Reading) z) )
WHERE [TagId] AND x.AreaId != y.AreaId AND x.AreaId = z.AreaId
WITHIN 1 minuteRETURNx.TagId, x.ProdId, x.AreaId, y.AreaId,
retrieveHistOfMvmt(x.TagId)
Stock Market AnalysisEvent SEQ(Stock+ a[], Stock b)Where skip_till_next_match(a[]!,b) { [symbol] AND a[1].volume > 1000 AND
a[i].price > avg(a[…i-1].price)) AND b.volume < 80%*a[a.LEN].volume }
Within 1 hourReturn a[1].symbol, a[].(price,volume), b.
(price,volume)
Medical ComplianceEvent seq(MEDICINE-TAKEN x, MEDICINE-
TAKEN y) Where [name=‘John’] [medicine=‘Antibiotics’]
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 14
Monitoring & Anomaly Detection
Alert if a bottle of drug with an unknown manufacturer appeared in the supply chain.
Alert if a bottle of drug with an unknown manufacturer appeared in the supply chain.
Alert if a case of vaccines has not been seen at three consecutive points in the supply chain.
Alert if a case of vaccines has not been seen at three consecutive points in the supply chain.
Alert if a nominally peanut-free bottle was filled with food containing peanuts. Alert if a nominally peanut-free bottle was filled with food containing peanuts.
Alert if two credit cards with the same number have been found in different locations.
Alert if two credit cards with the same number have been found in different locations.
Alert if a patient has taken overdoses of medicine in past 12 hours. Alert if a patient has taken overdoses of medicine in past 12 hours. ……
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 15
Receiving AreaShelf 1
Shelf 2
Shipping Area
RFID Simulator: Supply RFID Simulator: Supply ChainChain
UUNIVERSITY OF NIVERSITY OF MMASSACHUSETTSASSACHUSETTS, A, AMHERST • MHERST • Department of Computer Science Department of Computer Science 16