Experiment 1: Lab

The setup of our first experiment, conducted in our lab, is shown in Fig. \ref{FigSim1Setup}. The hardware used in the experiment is shown in Table \ref{TableSim1HW}. As noted in Table \ref{TableSim1HW}, $R_1$ and $R_2$ were identical systems for Case 1 and all systems were running Fedora Core 6. The laptops were configured in Ad-Hoc mode using channel 1 with no overlapping wireless networks present. All cards were rate locked at 54 Mb/s for their transmission rate. Additionally, all systems and wireless cards had their power management features disabled. Finally, all systems were connected to power such that the battery would not play a role in packet reception \cite{Zhao:SenSys2003}.

In the experiment, nodes $R_1$ and $R_2$ received a broadcast stream from Node $S$. Node $R_m$ was used to record all wireless traffic on the channel, including management frames via the MadWifi? driver. Node $S$ broadcasted a UDP, CBR stream of 16 Mb/s\footnote{16 Mb/s was the maximum achievable rate of our equipment, before significant loss was seen due to send buffer overflow.} with 1500 byte data packets (about 1400 packets per second) to both receivers and the monitoring node. Each experiment was performed several times, 15 minutes each, in order to verify results\footnote{The original data in addition to further analysis and results can be viewed at \emph{http://netscale.cse.nd.edu/twiki/bin/view/Main/WirelessReliability}.}.

-- DavidSalyers - 14 Aug 2007

• setupTable.png:
  

• Exp1LossRate? .png:
  

• Exp1LossCorrelation? .png:
  

• Exp1BeaconCorrelation? .png:
  

• Exp1Layout.eps: Exp1Layout? .eps