36x60 poster template

1
A Thorough Link-Level Study of 802.11g Channel
Dynamics Allen Ott, Joe Davis, Shaoen Wu, and
Dr. Saad Biaz AUBURN UNIVERSITY
Testing Materials and Methodology
Preliminary Results
Problem and Background
Indoor Shown below are plots of the RSSI vs.
Time for Location 1 in our experiments. The RSSI
was recorded for each packet by the TCPDUMP
program we used. This graph shows how the RSSI
varied over the sixty seconds of the average four
runs for each data rate. Outdoor Shown below
are plots of the delivery rate vs. time for
Location 2 in our experiments. Delivery rate was
computed by counting the number of packets lost.
This amount was found by recording the sequence
number contained in the IEEE 802.11 packet
header. Public Site Our preliminary results
show that only 48.2 of packets were actual data,
while 39 were used for management and 12.8 for
control.
Test Cases We ran a total of 9 different
experiment. Each case consisted of a different
location and a different data rate. The cases
were as follows -Location 1 -Location
2 Location 3 Location 4 -54 Mb/s -54
Mb/s -24 Mb/s -6 Mb/s -24 Mb/s -24 Mb/s -
6 Mb/s - 6 Mb/s - 6 Mb/s Hardware We
used very common hardware for our experiments.
We used a three laptop computers and a wireless
access point. Two of the laptops were IBM Lenovo
T60's and one was a system76 Darter laptop. We
used a Belkin wireless g network router for our
access point. One of the T60 laptops were used
as the mobile wireless client. This laptop used a
pcmcia Linksys wireless G adapter for its
wireless connection. The other T60 laptop was
used to record the traffic coming into the access
point. Its wireless card was placed in monitor
mode so that it could not cause any interference
in our tests. The system76 machine was used as
the wired client that was connected to the access
point by Ethernet. Software Free, open-source
software was used throughout our testing. Linux
distributions that ran the 6.2.25 kernel were
used as our operating systems. The T60 laptops
used Fedora Core 9 and the system76 used Ubuntu
8.04. The network sniffer laptop used a slightly
modified version of TCPDUMP to intercept and
record the wireless traffic. Our modifications
customized TCPDUMP's output to better accommodate
our analysis. Iperf, a network performance
utility, was used to generate traffic between the
two clients in our experiment. The ath5k
wireless driver was used in our wireless client
and network sniffer. Test Method In both the
indoor and outdoor cases, our clients were
configured as shown below in Figure 1. UDP
traffic flowed from the wireless client client,
through the access point, to the wired client.
Wireless packets were recorded by the sniffer for
a 60s time period. Each test case was run 6times.
Figure 2 shows the general layout of our indoor
test environment. The outdoor tests were
conducted in an open field with all hardware
elevated approximately 4ft to minimize the
destabilizing effects of ground reflection. The
public site field tests were conducted on Auburn
University's wireless campus network. The network
sniffer was placed in close proximity to an
access point and recorded all traffic to and from
that access point over a 1 hour time period. MAC
address filtering was used to ensure that no
foreign traffic was recorded.

• A research gap exists in the area of wireless
  channel dynamics. With a wired connection, like
  Ethernet, the channel quality is very stable. The
  performance of the wired link is degraded
  primarily by collisions. With a wireless link
  something as basic as fluctuations in the air
  quality can adversely affect the quality of the
  wireless channel. Since wireless networks are
  quickly becoming an ubiquitous part of our lives,
  the lack of understanding of the effects of
  wireless channel dynamics on network performance
  is an important and pressing problem.
• We recognize that their have been several past
  studies in this area. However, in general, we
  believe these studies were limited in one or all
  of the following ways
• Most studies are done on 802.11b, which is an
  older technology. With the standard being
  802.11g and 802.11n quickly on the rise, 802.11b
  is not relevant in most cases these days.
• They measurements are collected on limited test
  beds, or they are performed on unrealistic test
  beds, or a combination of both.
• The studies also focused on an overall
  performance evaluation and did not analyze the
  specific effects of channel dynamics on a
  detailed level.

Outdoor Test Results
Indoor Test Results
Purpose

• The purpose of this study is to overcome the
  limitations of past studies and to inform the
  design of future wireless technologies of the
  effects of channel dynamics. Our study achieves
  this purpose by
• Using the IEEE 802.11g wireless standard
  exclusively.
• Testing in 3 different environments, two
  controlled and one field test. We perform our
  experiments on an indoor test bed, an outdoor
  test bed, as well as the Auburn University
  wireless campus network.
• Focusing on packet by packet analysis instead of
  an overall network performance analysis. A packet
  by packet analysis can give much better insight
  into how the channel fluctuates between small
  periods of time.

Preliminary Conclusions

• In public access points, we confirm that much of
  the available bandwidth is used for overhead
• From our outdoor tests, we find that while the
  54Mbps link is extremely unstable it generally
  outperforms the more stable 24Mbps link. The
  6MBbs is the highest quality and most stable as
  expected.
• The indoor test results show that the RSSI of the
  6Mbps data rate is the least stable of the three
  data rates tested.
• Also the indoor results show, that as the data
  rate increases the stability of connection signal
  increases.

Figure 2 Indoor Layout
Figure 1 Experimental Setup