Check out the new USENIX Web site.


Design Overview

Figure 1 is a high-level illustration of the DenseAP system. The system consists of several DenseAP nodes (DAPs) which provide wireless service and a DenseAP controller (DC) which manages the DAPs.

Figure 1: Overall architecture of the DenseAP system.
\begin{figure}\begin{center}
\epsfig{figure=figs/new_architecture3.eps, width=0.7\columnwidth}
\end{center} \vspace{-0.5em}
\end{figure}

A DAP is a programmable Wi-Fi AP connected to the wired network. Each DAP periodically sends summaries to the DenseAP controller comprising of a list of associated clients, their traffic pattern summaries, RSSI values of a few packet samples from their transmissions, current channel conditions, and reports of new clients requesting service from the network. We classify DAPs into two categories: we refer to DAPs that do not have any clients associated with them as passive; those that have at least one associated client are called active.

The DC manages the DAPs. The periodic reports sent by the DAPs provide the DC with a global view of the network activity. Using this global view, the DC selects the right DAP for a client, allocates channels to DAPs, performs load balancing when needed, handles client mobility, and deals with DAP failures.

In the next section, we describe the mechanisms the DC uses to ensure that the client associates with the selected DAP. Then, we describe the algorithms involved in selecting the DAP and a channel for the DAP. We discuss power control and related issues in Section 5

NSDI-2008