Transcript Analysis of a Campus wide Wireless Network

Topic on WLANS
• IEEE-802.11 (Hao Lian)
• Analysis of campus wireless
network(Ao Shen)
• Comparison between 3G and WiFi(Bichen Wang, Chen Chen)
Standardization of Wireless Networks
• Wireless networks are standardized by IEEE
• Under 802 LAN MAN standards committee.
Application
Presentation
ISO
Session
IEEE 802
OSI
standards
7-layer
Transport
model
Network
Logical Link Control
Data Link
Physical
Medium Access (MAC)
Physical (PHY)
IEEE 802.11 Overview
Goals
•To deliver services in wired networks
•To achieve high throughput
•To achieve highly reliable data delivery
•To achieve continuous network connection.
• Adopted in 1997.
Defines;
• MAC sublayer
• MAC management
protocols and services
• Physical (PHY) layers
– IR
– FHSS
– DSSS
Components
• Station
• BSS - Basic Service Set
– IBSS : Infrastructure BSS : QBSS
• ESS - Extended Service Set
– A set of infrastrucute BSSs.
– Connection of APs
– Tracking of mobility
• DS – Distribution System
– AP communicates with another
Wireless connection
WLAN 802.11 network
AP
STA
STA
DS (usually Ethernet)
AP
STA
STA
STA
STA
STA
BSS
BSS
ESS
Services
• Station services:
– authentication,
– de-authentication,
– privacy,
– delivery of data
• Distribution Services ( A thin layer between MAC and
LLC sublayer)
–
–
–
–
–
association
disassociation
reassociation
distribution
Integration
A station maintain two variables:
• authentication state (=> 1)
• association state
(<= 1)
Services example : Roaming
Laptop computer
1- Authenticate and
associate
2
Laptop computer
2 – Laptop roaming
3 – Authenticate (if needed)
and (re)associate
1
3
4 – Notify the new location
of the laptop (disassociation
of AP1)
4
AP1
IEEE 802.11 overview
AP2
AP3
Services example : “Out of service”
Old BSS
Laptop computer
New
BSS
New
BSS
Laptop computer
Laptop computer
AP2 is
out of
service
AP1
IEEE 802.11 overview
AP2
AP3
Medium Access Control
Deals:
• Noisy and unreliable medium
• Frame exchange protocol - ACK
• Overhead to IEEE 802.3 • Hidden Node Problem – RTS/CTS
• Participation of all stations
• Reaction to every frame
MAC functionalities
• Reliability of data delivery service
• Control of shared WL network
– Distributed Coordination Function (DCF)
– Point Coordination Function (PCF)
• Frame Types (informational section)
• Management
• Privacy service (Wired Equivalent Privacy - WEP)
IEEE 802.11 overview
DCF Operation
• Carrier Sense Multiple Access Collision Avoidance
(CSMA/CA), uses binary exponential back off (Same
as in IEEE 802.3)
• IEEE 802.3 use collision detection algorithm.
• IEEE 802.11 use collision avoidance (CA) algorithm
• Listen Before Talk – LBT (don’t transmit while others
transmit to avoid collision)
• Network Allocation Vector (NAV) – the time till the
network will be cleared from any transmitting.
• The NAV with the LBT assist to avoid collisions (CA)
IEEE 802.11 overview
DCF Operation
PCF Operation
• Poll – eliminates contention
• PC – Point Coordinator
– Polling List
– Over DCF
– PIFS
• CFP – Contention Free Period
– Alternate with DCF
• Periodic Beacon – contains length of CFP
• CF-Poll – Contention Free Poll
• NAV prevents during CFP
• CF-End – resets NAV
Other MAC Operations
• Fragmentation
– Sequence control field
– In burst
– Medium is reserved
– NAV is updated by ACK

Privacy
 WEP bit set when
encrypted.
 Only the frame body.
 Medium is reserved
 NAV is updated by ACK
 Symmetric variable key

WEP Details
 Two mechanism
 Default keys
 Key mapping
 WEP header and trailer
 KEYID in header
 ICV in trailer
 dot11UndecryptableCount
 Indicates an attack.
 dot11ICVErrorCount
 Attack to determine a
key is in progress.
MAC Management
• Interference by users that have no concept of data
communication. Ex: Microwave
• Interference by other WLANs
• Security of data
• Mobility
• Power Management
Analysis of a Campus wide
Wireless Network
• Introduction
• Trace Collection
• Results
Introduction
• Research time: 2001 Fall
• Research place: Dartmouth(161
buldings, 476 APs)
• Research Target: Wireless network
analysis in Dartmouth.
• Note: The paper only applied to
Dartmouth 2001.
Trace Collection
•
•
•
•
Syslog
SNMP
Sniffers
Other important definitions
Syslog
• Contains:
★AP name
★MAC address of card
★Message time
★Message type
Message Type of Syslog
• Associated
A card selects one AP
• Roamed
A card changes its current AP to another
• Deassociated
A card disconnects with one AP
SNMP
• A kind of heart-beat message, poll
per 5 minutes
• Contains:
★MAC address of card
★Inbound bytes
★Outbound bytes
Sniffers
• An application which collects
packets and by extracting packets’
header, we can analyze more
detailed information about users,
such as types of packets,
application-layer protocol used.
Sniffers
• We assemble “Sniffers” in four
buildings
★Sudikoff(6 APs)
★Brown(2 APs)
★Berry(13 APs)
★Collis/Thayer(9 APs)
Other Important Definition
• Roamer Card
• Mobile Card
• Session
Session
• Starts when a card associates with
an AP.
• Ends:
★Changes one AP to another
★Network Problem: Power Off .etc
Results & Analysis
•
•
•
•
•
Traffic
Card
Session
AP
Protocol
Traffic
• The busiest card transferred 117GB, while the median card
transferred only 350MB. On the busiest day, the traffic has the
amount of 240GB, while the median daily traffic is only 53MB.
Inbound traffic is greater than outbound traffic.
Traffic
•
•
•
•
•
77 days’ traffic overview
Weekly pattern
Reduction in Thanksgiving
Reduction in the end
Holes
Traffic
Based on Week
• Monday is busiest,
because weekday starts
with Monday
• Friday and Saturday are
quietest, because
students always rest on
that two days
• From Sunday, the traffic
is beginning to grow,
because students usually
start to finish homework
on Sunday
Graph
Traffic
Day Pattern
• Around 10:00 AM, busiest
• In the afternoon, very steady
• After 12:00 AM, is declining
Graph
Card
• Activity varies significantly
from active only once to
active all 77 days(1706
cards)
• Median activity days: 28
days
Card
• 77 days’ overview of
the number of active
cards
• Follow the pattern of
Traffic
Card
• Daily Card Activity
• Most active in the
afternoon, very steady in
the afternoon
• From 12AM, huge
reduction
Session
• Median Session length: 16.6 min
• 71% less than one hour
• 27% less than one minute(Overlap AP areas)
Roamed Session
• 18% of all the sessions
are roamed session
• 60% of the roaming
sessions roamed only
within only one subnet
AP
• 476 APs, more than 20APs not found in trace
• Each day between 171 and 352 APs are used
AP
• AP traffic:
• Busiest AP: 2GB per day
• Median: 39 MB
Protocol
•
•
•
•
Get protocol by packet header and port number
99% are IP packets
In all of the IP packets: 99% are using TCP or UDP
Application-layer protocol: http(53%), dantz(15%)
…
Protocol
• All traffic is concerned with
web browsing, email,
backup, file transfer, and file
sharing
• Inbound traffic is more than
outbound traffic
Protocol
• Although it is assymmetric
in terms of traffic, it is
symmetric when it comes
to the number of TCP
connections.
• Inbound connections
equal to outbound
connections
• Performance Comparison of
3G and Metro-Scale WiFi for
Vehicular Network Access
• Pralhad Deshpande, Xiaoxiao Hou, and Samir R.
Das
• Computer Science Department, Stony Brook
University
• Stony Brook, NY 11794, USA
INTRODUCTION
• 3G:
– licensed bands and macrocells with large coverage
areas.
– base station and associated radio access network
setup have significant capital and operational costs.
– ubiquitous.
• WiFi:
– unlicensed spectrum.
– access point (AP) coverage is relatively smaller and
typically capital and operational costs are lower.
– free or inexpensive, and would provide a significantly
higher bit rate
– local area
PROBLEM
• Can WiFi be used effectively in outdoors and
mobile scenarios to reduce the load on the
expensive 3G networks?
MEASUREMENT SETUP
• Network
– Optimum WiFi provided by Cablevision and has roughly 18,000
APs
– Verizon’s EVDO Rev 3G access
• Testbed
– Dell Latitude laptop running Linux as the client in the car
– carrier-grade interface and transmit power choices + Verizon’s
USB-based USB760 EVDO Rev
– DHCP
– TCP maximum retransmission timer on the server side is 1 sec.
ps: measure end-to-end throughput only instead of the throughput on the wireless hop because
of lack of access to the provider network
• Driving Scenarios
– Long Drive (Once):
• 500 miles
• vehicle speed varied depending on the road traffic
• reasonable sample of the quality of WiFi access from moving
vehicles in a metro-scale deployment scenario.
– Short Repeated Drives (10 times):
• 9 mile
• selected stretch where the quality of the AP coverage is
good.
• experimental results.
• Log
– TCP throughputs/second (in-staneous throughputs) on both the
connections along with GPS location and vehicle speed.
MEASUREMENT RESULTS
• Quality of WiFi Coverage
• CDF of run lengths (consecutive 1 sec segments) with zero and
non-zero throughputs seen on WiFi.
• ComparingWiFi and 3G Throughputs
• CDF of instantaneous TCP throughputs for WiFi and 3G
• CDF of relative difference of instantaneous throughputs (in Kbps)
between WiFi and 3G. Plot for the long drive only.
• Correlation with Vehicle Speed
Short Drive
Long Drive
• Correlation with Location
• Comparison of total and location entropies for 3G and WiFi
networks
• H(X|li) is the entropy of throughput for a specific location li.
• Temporal Correlation
• Autocorrelation R(k) of the instantaneous throughputs measured in
1 sec intervals
CONCLUSION
• WiFi
– frequent disconnections even in a commercially operated, metroscale deployment;
– when connected indeed delivers high throughout even in a
mobile scenario.
• 3G network
– much lower throughputs
– much better coverage and less throughput variability.
• A hybrid design that exploits the best properties of the
two networks opportunistically can be very successful.
Better throughput + lower cost for the provider by moving
expensive 3G bits onto WiFi networks.
Thanks so much!