SLIDE 1
The Changing Usage of a Mature Campus-wide Wireless Network
CS525m – Mobile and Ubiquitous Computing
Andrew Stone
SLIDE 2 Worcester Polytechnic Institute 2
Overview
- Project Goal
- Data Collection Methods
- Findings and Data Analysis
- Conclusions and Recommendations
SLIDE 3 Worcester Polytechnic Institute 3
Project Goal
- Compare WLAN network usage
statistics at Dartmouth college from fall 2001 and 2003/2004
– Number and type of devices – Applications in use
- Better understand user behavior
SLIDE 4 Worcester Polytechnic Institute 4
Networked Devices
– PBX converted to VoIP in 2003 – Softphones and VoIP phones
– Laptops, PDAs, phones, etc identified by using tcpdump traces and an OS fingerprinting tool (p0f)
SLIDE 5
Worcester Polytechnic Institute 5
Networked Devices
SLIDE 6 Worcester Polytechnic Institute 6
Data Collection (2001)
- Analysis of a Campus-wide Wireless Network (2001)
http://users.wpi.edu/~astone/mobilecomputing/kotzwlananalysis.pdf
- 476 802.11b APs over 161 buildings
– 430 were tracked
- 115 Subnets
- Syslog tracking started before
– Contains authentication, association, and roaming information
- AP SNMP Polling for Client specific
counters
– MAC, IP, Signal Strength, traffic info
SLIDE 7 Worcester Polytechnic Institute 7
Data Collection (2003)
- 566 APs over 188 buildings
- <115 Subnets (started using VLANs)
- >75% of undergrads own laptops
- VOIP phone records
- 18 sniffers covering 121 APs (most
popular areas from 2001)
SLIDE 8 Worcester Polytechnic Institute 8
Data Collection
– Over 1700 different wireless cards – 11 week period
– Over 7000 different wireless cards – 17 week period
SLIDE 9 Worcester Polytechnic Institute 9
Definitions
- Card, Session, Active card, Active
AP, Roam, Roaming Session, Roamer Card, Inbound, Outbound
- Mobile session, Mobile card
SLIDE 10 Worcester Polytechnic Institute 10
Mobile Session
Distance between any two APs in a session > ab, bc, cd, da, ac, or bd
SLIDE 11 Worcester Polytechnic Institute 11
Questions asked
- Has the population grown?
- Have usage patterns changed?
- Where do users visit?
SLIDE 12
Worcester Polytechnic Institute 12
Has the population grown?
SLIDE 13
Worcester Polytechnic Institute 13
Active cards per day (2001)
SLIDE 14
Worcester Polytechnic Institute 14
Active cards per day (2003)
SLIDE 15
Worcester Polytechnic Institute 15
Active cards per hour (2001)
SLIDE 16
Worcester Polytechnic Institute 16
Active cards per hour (2003)
SLIDE 17
Worcester Polytechnic Institute 17
Heavy User Distribution
SLIDE 18
Worcester Polytechnic Institute 18
Average active days per week per user
SLIDE 19
Worcester Polytechnic Institute 19
Average active days per week per user
SLIDE 20
Worcester Polytechnic Institute 20
Active APs per day
SLIDE 21
Worcester Polytechnic Institute 21
Active APs per day
SLIDE 22
Worcester Polytechnic Institute 22
Average Active cards per active AP per day
SLIDE 23
Worcester Polytechnic Institute 23
Average Active cards per active AP per day
SLIDE 24
Worcester Polytechnic Institute 24
Max cards per hour, for busiest buildings
SLIDE 25
Worcester Polytechnic Institute 25
Max cards per hour, for busiest buildings
SLIDE 26
Worcester Polytechnic Institute 26
Daily Traffic (GB)
SLIDE 27
Worcester Polytechnic Institute 27
Daily Traffic (GB)
SLIDE 28
Worcester Polytechnic Institute 28
SLIDE 29
Worcester Polytechnic Institute 29
Total traffic
SLIDE 30
Worcester Polytechnic Institute 30
Total Traffic
SLIDE 31
Worcester Polytechnic Institute 31
Near/Far Traffic
SLIDE 32
Worcester Polytechnic Institute 32
Near/Far Traffic
SLIDE 33
Worcester Polytechnic Institute 33
Average daily traffic per AP, by AP category
SLIDE 34
Worcester Polytechnic Institute 34
Average daily traffic per AP, by AP category
SLIDE 35
Worcester Polytechnic Institute 35
VoIP calls by hour
SLIDE 36
Worcester Polytechnic Institute 36
VoIP CCDF of call duration
SLIDE 37
Worcester Polytechnic Institute 37
P2P Total Traffic
SLIDE 38
Worcester Polytechnic Institute 38
P2P Near/Far Traffic
SLIDE 39
Worcester Polytechnic Institute 39
Streaming Media Total Traffic
SLIDE 40
Worcester Polytechnic Institute 40
Streaming Media Near/Far Traffic
SLIDE 41
Worcester Polytechnic Institute 41
Mobility Time spent at home location
SLIDE 42
Worcester Polytechnic Institute 42
Mobile Session duration
SLIDE 43
Worcester Polytechnic Institute 43
Mobility
Session diameter, distribution across sessions, by application
SLIDE 44 Worcester Polytechnic Institute 44
Conclusions and Recommendations
- Found dramatic increase in usage and change in
applications
- Because the majority of users stay in their home
location network caching and prediction-based schemes may be beneficial
- Device profiling: single VLAN campus wide for
VoIP and PDA devices and building subnets for laptops
- Compare wireless usage against wired usage
SLIDE 45 Worcester Polytechnic Institute 45
Future Work
- WLAN moving to 802.11a/b/g
- Campus Cable TV network migrating
to IP-based streaming video
SLIDE 46
Worcester Polytechnic Institute 46
Thoughts and Questions?