Architecture, Applications, and Data Analysis of a Wireless Network Location Service

1
Architecture, Applications, and Data Analysis of
a Wireless Network Location Service
Thesis Defense
April 11th, 2006

• Jeremy Shaffer
• Electrical and Computer Engineering Department
• Carnegie Mellon University

2
Outline

• Contributions
• Motivation
• Mobility
• Locator_at_CMU
• Toolkit
• Applications
• Contact Disease Tracking
• Unknowing Bystander
• Future Prediction
• Conclusions
• Questions

dove?
Where?
Donde?
waar?
????
Wo?
onde?
où?
3
Main Contributions

• Defining wireless user mobility on 802.11 network
• Analysis of location service architectures
• Detailed model and implementation of centralized
  wireless location service
• Creation of toolkit for location-based
  applications
• Implementation of location based applications
  that demonstrate end-to-end capabilities

4
Motivation

• Strong need for reliable, scalable location
  service for applications
• Current techniques do not support large scale
  implementations, but are focused on single
  devices or small areas
• Lack of understanding on user movement patterns
• Disorganized approach to location services

5
Approach

• Utilize 802.11 Wireless Network infrastructure
  (Access Points)
• Research divided into three main serial phases
• Mobility/Access Point Usage Analysis
• Locator_at_CMU Service and Toolkit/API
• Location-Enabled Applications Implementation

6
Mobility on Wireless Network

• Detailed study done in 2002-2003 also compared to
  new trace in 2005
• Found much larger number of users on the system
• Basic movement and mobility changed little over 2
  years
• Data did show a small, highly mobile group
  emerging
• Answered additional questions such as
• - How can users be classified by movement
  history (home/favorite site classifications)?
• - Do factors such as weather, day of week, and
  time of day effect mobility?

7
Mobility on Wireless Network
8
Architecture Comparison

• Examined four of the main types of location
  architectures (Centralized Push, Centralized
  Pull, Distributed Push, Distributed Pull)
• Selected location aware messaging as a
  representative application
• Quantified data flow requirementsand messages
  for location based application
• Showed Locator_at_CMUs centralized
• pull model performs better than distributed for
  location aware messaging

9
Architecture Comparison
10
Architecture Comparison
Comparison of Architectures using Instant
Messaging Application Polling/Push Frequency 1
min 3,000 wireless clients 48 Buddies per User
11
Locator_at_CMU

• Implemented on Wireless Andrew
• 1,000 APs
• 5,000 peak concurrent users
• Centralized-Pull architecture using relational
  database
• Provides omniscient view of network usage
• Modularized by major components for scalability
• Web Interface (Registration and Rules Creation)
• Database
• Tables
• Stored Procedures
• Polling Logic
• Toolkit

12
Major Components
13
Implementation
14
Access Point Data Polling and Storage
Step 1) Sending Request

• Two Methods for Database Storage
• Original (1 row per MAC per Poll)
• Modified (Row only per AP move)
• Modified offers 94 savings over original

Step 2) Saving to Database
Step 3) Getting Next AP Info.
15
Queueing - Reactive
M / M / 1 / 8 / 8 / FIFO
Number of Client Application Requests Per Minute
Number of Client Application Requests Per Minute
Example 1 Five groups are trying to find a
conference room with no one it. (25 conference
room locations x 5 requestors 125
requests) Example 2 E-Coupons Application
running in University Center. Example 3 All GSIA
users running Location Aware Messaging
16
Clustering
17
Toolkit

• Over 20 location-based apps/services chosen as
  basis to determine Toolkit coverage
• Full description and grouping of all toolkit
  functions/data types
• Matrix demonstrates relations between functions
  and applications
• Java classes written 85 different functions,
  over 6,000 lines of code.

18
Security

• Developing trust in the system is critical
• Potential vulnerabilities and solutions
  identified
• Security mechanisms implemented
• SSL
• Double Registration
• System Configuration
• Passwords

19
Privacy

• Privacy handled by rules based approach
• Rules Creation module via web-page
• Rules Processor filters results returned
• Rules can be applied to all applications

Karnaugh Map Representation
20
Location-Based Applications

• Complete/Compound Applications
• Where are users now and where have they been
  (past/present)
• Contact/Spread Tracking
• Where were users (past)
• Unknowing Bystander Service
• Where will users be (future)
• Crowd Predictor
• Focused Applications
• Mobility Analysis (Used for Chapter 3)
• Location Aware Instant Messaging (Used for
  Chapter 4)
• Built by Others
• Linking with Aura
• Other Student Projects

21
Applications Contact/Spread Tracking

• Scenario Someone is infected how many people
  do they spread the disease too in various
  situations?
• Divided wireless users into infecting agents and
  general users.
• Selected 10 infecting agents (4 undergrads, 2
  grads, 2 faculty, and 2 staff)

22
ApplicationsContact/Spread Tracking

• Each user given a wireless PDA to carry around
  for 24 hour trace.
• Applied four different infection scenarios
• Direct-Primary (Direct Contact with Infecting
  Agent)
• Direct-Any (Direct Contact with Anyone Infected
  Can Propagate)
• Direct-Indirect-Primary (Contact with Infecting
  Agent plus 1hr. Residual)
• Direct-Indirect-Any (Contact with Anyone Infected
  or their 1hr. Residual Can Propagate)

23
Contact/Spread TrackingDirect-Primary
24
Contact/Spread TrackingDirect-Indirect-Primary
25
Contact/Spread TrackingDirect-Any
26
Contact/Spread TrackingDirect-Indirect-Any
27
Contact/Spread TrackingSummary
28
Contact/Spread TrackingPerformance
29
Contact/Spread TrackingPerformance
30
Applications Unknowing Bystander

• Scenario.. An event happens at location X.
  People nearby might not even be aware but can
  have valuable information.
• Event Examples
• Crime (Burglary/Theft/Murder, etc.)
• Lost item, pet, or person
• Possible Users
• Police, Homeland Security (Citizen Watch Corps)
• Individuals
• Used campus crime data to determine how many
  network users near area and could be potential
  witnesses
• What percent of time would there be a potential
  witnesses?

31
Unknowing BystanderResults

• 15 of the 16 crimes had potential witnesses
• Average value of 12.8 for potential witnesses
• Median value of 4.5 for potential witnesses
• Chance of at least 1 witness for 4.5 witnesses
  with likelihood of 5 (21) 10 (38) 33 (83)

32
ApplicationsCrowd Predictor

• Use information from historical data to populate
  an application to predict future crowds at a
  location (Neural Network)
• Can be used by organizations to find best spot to
  setup table
• Allow for other limited criteria (such as type of
  space, time of day, day of week)

33
Applications Crowd Predictor - Metrics

• Select 12 different APs
• 3 in Classroom Areas
• 3 in Office Areas
• 3 in Public Areas
• 3 in Dorm Areas
• Predict wireless crowds at 12 test APs at 5
  different day/time combinations and compare to
  observed results
• Look at effect of time of day, day, and type of
  area predicting
• Measure time taken to run prediction routines for
  each

34
Applications Crowd Predictor
CPU Execution Time 2.01 -2.91 sec. Data Size
180KB 2wks 90KB 1 wk
35
Conclusions

• Locator_at_CMU architecture is an efficient and
  scalable method for obtaining location data on
  wireless devices
• Provides an omniscient view and
  device-independent view for large scale
  implementation
• The system supports a wide variety of
  location-based applications
• Can be implemented on any AP based network even
  much larger ones

36
Papers from Thesis

• 4 Conference papers accepted/presented on system
• 1 Book Chapter Written
• 2 Conference papers submitted pending review
• 2 Conference papers to be submitted within next
  month
• 1 Journal Article to be submitted

37
Questions?