Establishing the Baseline

We begin by evaluating the performance of the corporate WLAN, to provide a baseline against which we can compare the performance of our system.

As mentioned earlier, the testbed area is served by a single corporate AP. To establish the baseline performance, we had a group of clients associate with the corporate AP. The clients then simultaneously carried out a one minute TCP download from a server on the wired network. We varied the group size (the number of clients) from 2 to 12. The experiment is repeated 10 times for each group size. Each time, the group members are selected at random from among the available clients. We performed similar experiments for upload.

The results of this experiment are shown in Figures 8 and 9. Each point represents the median per-client throughput, and the error bars show SIQR.

For both 802.11a and 802.11g, the median per-client throughput drops as number of simultaneously active clients increases. However, the 802.11g numbers are substantially lower than the 802.11a numbers. This is because in 802.11g mode, the corporate AP sends out a CTS-to-self before every packet to avoid interfering with 802.11b clients. This is a well-known, and well-studied issue. Since we have not implemented the CTS-to-self feature for DAPs, we will refrain from directly comparing the performance of DenseAP and the corporate network in 802.11g mode.

Figure 8: Baseline performance: 802.11a

Figure 9: Baseline performance: 802.11g