In this section, we discuss in detail i) whether rate diversity exists in today's 802.11b networks and ii) whether a single user or multiple users are actively exchanging data during the intervals in which the network is saturated.
To investigate the prevalence of rate diversity, we collected traces of wireless network traffic at one-day Iris student workshop at MIT. There were about 45 attendees and more than half turned on their wireless laptops. We set up a laptop to sniff data during each of the three 90-minute sessions, WS-1, WS-2 and WS-3, all of which took place in a single room of about 40' x 25'.
Figure 1 shows the fractions of data bytes
transferred using each of the four possible rates during each
session. It is clear that rate diversity exists even in a relatively
small room. During WS-2, more than 
of the data bytes were
transferred using data rates lower than 11 Mbps.
We also set up an experiment to investigate how an AP change data rates
to various clients in indoor office environments. We placed a Cabletron
Roamabout-2000 AP in a 18' x 14' office 7' above ground. A
sender with a wired connection to the AP sent unicast UDP data packets
at the saturation rate simultaneously to four different receivers.
The first node was about 4' away from
the AP, the second 12' and one thin, wooden wall away, the third 26'
and two thin wooden walls away and the fourth 30' and two thick walls
in between.
As shown in Figure 1 (see EXP-1), more than 
of the bytes were transferred using the lowest data rate.
In fact, a recent extensive wireless network usage study on a university campus has found that the average received signal strength varies widely even among positions that are within 20' of an access point [19]. Thus, we believe that rate diversity is prevalent in many indoor WLANs and its impact would be much more pronounced with mixed deployments of 802.11b and 802.11g networks.
The negative impact of rate diversity is significant only if the following two conditions are true: i) more than one competing node exchange data during the periods in which network is saturated and ii) competing nodes use diverse data rates. Our analysis of this particular workshop trace data, however, shows that the network is well over-provisioned with 7 APs providing a combined channel capacity of 33 Mbps. However, recent studies have shown that in many enterprise networks [2] and university residential halls [18], WLANs carry significant traffic and contain many APs that have a lot of busy or congested periods. In the next subsection, we show that more than one user competes for channel access during those congested periods.