Outline

1
(No Transcript)
2
Outline

• Introduction to WLAN
• WLAN Monitoring
• Related Work
• Project Goals

3
Current State of the Art
GEO
MEOs
Space Segment
LEOs
Ground Segment
Wireless Segment

Ad Hoc Sensor Networks
WirelessLANs
2G
4
Current State of the Art
GEO
MEOs
Space Segment
LEOs
Ground Segment
Wireless Segment

Ad Hoc Sensor Networks
WirelessLANs
2G
5
Wireless LANs (a.k.a. Wi-Fi)

• Development of IEEE 802.11
• High data transfer rate
• Ease of installation
• Anytime, anywhere connection
• Extremely large application areas
• Companies
• Airports
• Hospitals
• Coffee houses
• University Campuses

6
802.11 a vs. b vs. g
7
WLANs (Future)

• The number of public Wi-Fi hot spots will grow to
  167K by 2008, up from virtually nil in 2001
• Wireless public area access service revenue is
  expected to grow from 3.9 million in 2001 to
  224.7 million in 2005
• By the year 2010, the number of wireless data
  subscribers will hit 1B!

8
WLAN Monitoring

• Monitoring required for
• Pricing
• Maintenance
• Diagnostics
• Monitoring types
• Active (other groups are doing that!)
• Passive (our focus)
• Generic tool for WLAN analysis

9
Georgia Tech Local Area Wireless Network (LAWN)
10
(No Transcript)
11
Related Work

• Traditional LAN analysis tools
• ping, traceroute, tcpdump, SNMP queries, etc
• Generally two types of monitoring work
• User level
• Flow level

12
Dartmouth College
1 D. Kotz and K. Essien. Characterizing Usage
of a Campus-wide Wireless Network, in Proc. ACM
MobiCom 02, pp. 107-118, September, 2002.

• 161 buildings, 476 Access Points (APs), 2000
  hosts
• Three types of traces
• syslog data collection
• SNMP polling
• tcpdump data collection
• Over 12 weeks
• Collected data
• User distribution
• Application types
• Session duration
• Data rates
• Mobility statistics

13
Application Types
14
Average Daily Traffic (GB)
15
Number of Hosts/Day
16
Avg. Active Hosts/Hour
17
Stanford University
2 D. Tang and M. Baker, Analysis of a
Local-Area Wireless Network, in Proc. ACM
MobiCom00, pp. 1-10, August 2000.

• 1 building, 12 APs, 74 hosts
• Three types of traces
• syslog data collection
• SNMP polling
• tcpdump data collection
• 12 weeks
• Collected data
• User distribution
• Offered load
• Mobility statistics

18
Avg. Number of Hosts
19
Number of Hosts Changing Location
20
SIGCOMM Conference
3 A. Balachandran, et. al., Characterizing
User Behaviour and Network Performance in a
Public Wireless LAN, in Proc. ACM SIGMETRICS
02, pp. 195-205, June 2002.

• Conference venue, 4 APs, 195 hosts
• Two types of traces
• tcpdump data collection
• SNMP polling
• 3 days
• Collected data
• User distribution
• Data rates
• Session duration
• Application distribution
• Mobility statistics

21
Application Type (during the sessions)
22
IBM
4 M. Balazinska, P. Castro, Characterizing
Mobility and Network Usage in a Corporate
Wireless Local-Area Network, in Proc. MobiSys
03, San Francisco, California, May 2003.

• 3 buildings, 177 APs, 1366 Hosts
• One type of trace
• SNMP polling
• 4 weeks
• Collected data
• User distribution
• Load distribution
• Mobility statistics
• Data rates

23
Session Duration Distribution
24
Flow-based Analysis
5 X. Meng et. al., Characterizing flows in
large wireless data networks, in Proc. ACM
MobiCom 04, pp. 174-186, 2004.

• Comparative work using traces from 1, 2, 3,
  4
• Flow-based analysis
• Temporal evaluation
• Spatial location
• WLAN traffic can be closely modeled using known
  distribution models

25
Arrival Rate vs. Weibull Regression Model
26
Roaming Flows
27
Georgia Tech
6 Ron Hutchins, Ellen W. Zegura, "Measurements
from a campus wireless network", in Proc. IEEE
ICC 2002, vol. 25, no. 1, pp. 3161-3167, April
2002.

• 18 buildings, 109 access points, 444 hosts
• Three types of traces
• ping sent to authentication hosts
• TCP data collection
• SNMP polling
• 20 weeks
• Collected data
• User distribution
• Session duration
• Mobility statistics

28
Distribution of Session Duration
29
Number of Users/Day
30
University of Saskatchewan, Canada
7 D.Schwab and R.Bunt, Characterizing the use
of a campus wireless network, in Proc IEEE
INFOCOM 04, 2004.

• 40 buildings, 18 APs, 134 users
• Two types of traces
• EtherPeek trace
• LEAP authentication data
• 1 week
• Collected data
• User distribution
• Data rates
• Application type
• Mobility statistics

31
Application Types
32
Number of Auth. vs. Traffic Load
33
User Distribution
34
Mobility Statistics
35
What We Have

• Authentication Data

36
What We Have (cont.)

• Flow Data

37
What We Have (cont.)

• SNMP Data

38
Performance Metrics

• Average User Service Time
• Average time a user spends in LAWN
• Usage statistics according to time and location
• Average Number of Users/AP
• Average number of users connected to a specific
  AP
• System Load
• Total number of users connected to the LAWN

39
Performance Metrics (cont.)

• Network Utilization
• Percentage of available bandwidth used by LAWN
  users
• Bottleneck Analysis
• Over-utilized APs
• Service Rejection Rate
• Average number of users that are rejected due to
  high load

40
Questions?