Performance with Fewer DAPs

We repeated the experiment described in the previous section, but using only 12 of the 24 DAPs deployed in our testbed. The 12 DAPs were selected at random. We used all 8 channels. The results are shown in Figure 16. For comparison purposes, we have also included lines showing the baseline corporate network performance, and the DenseAP performance when using 24 DAPs and 8 channels. As expected, the results show that using fewer DAPs hurts performance.

These results demonstrate the following: (i) More DAPs are beneficial and (ii) the performance of the DenseAP system degrades gracefully, if DAPs were to fail. We have already described how the DC can detect that a DAP has failed, and re-assign its clients to other DAPs. (iii) Note that the performance of the DenseAP system with 12 nodes is similar to the performance of the client-driven approach (Figure 13), with 24 nodes! In other words, the association policy can deliver similar performance with only half as many DAPs.

Figure 16: Performance with fewer DAPs