Check out the new USENIX Web site. next up previous
Next: Hardware Platform Up: Method Previous: Site allocation

Setting up a Node

A typical network node setup consists of a number of antennas and a computer system. We use 2 or more directional antennas to connect to other backbone nodes and one omni-directional antenna for local client access (See figure 3). A PC or other system provides the routing and access-point functionality. A typical node setup can be found in figure 4. The home-built and partly commercial antennas are connected to a computing platform that can, but does not need to be connected to the local network at the site using wired Ethernet.
Figure 3: Antennas as mounted on a building (2 directional antennas and one omnidirectional antenna).
\begin{figure}\begin{centering}
\psfig{file=381.eps,width=8cm}\small\itshape\end {centering}
\end{figure}
Setting up a network node requires some real hard hardware work to get the antennas lined up and affixed to the building. As we are using directional and polarized antennas to prevent interference, the alignment of the antennas is fairly critical. Once the antennas are set up, connection to the node machine is done using low loss coaxial cable. As keeping cable losses to a minimum is important, the length of the cable should be minimized as each meter of antenna cable introduces a loss of approximately 0.25 dB. The node machine should be as close as possible to the antennas. When using a small embedded system (see section 2.9) the node computer can even be mounted outside on the antenna pole. In this case, special measures have to be taken to weatherproof the setup (e.g. like preventing condensation). This setup can use Power over Ethernet (PoE) [PowerF] [PowerS]. A single UTP network cable (where the length is not an issue provided that it is less than 100 meters) connects the antenna setup to the power supply (indoors) and the on-site local wired Ethernet if needed.

Before and after the installation of the node hardware a number of basic alignment, throughput, and reliability tests have to be run. These tests are used to be sure that the links in the set-up will operate as expected. The expectation values for these tests are derived from the site survey and simulation results as described in section 2.4. The alignment test comprises of monitoring the signal strength and noise figure when a backbone link is established. By physically varying the antenna direction the signal as seen by the driver software on the wireless interface should be maximized, while the reported noise is minimized. Throughput and reliability is tested by transferring large chunks of data from one node to its neighbor and observing the packet loss and transfer rate.

Often, once the node is set up it is quite difficult to physically access the machine and the antennas because they mostly are located at remote locations and in buildings with complex access procedures. A typical test for the reliability is to copy some video data (often comprised of large files) through the node. This test differs from the test mentioned above where a single wireless link is tested, here the complete (routing) functionality of a node is tested. A test protocol is used to assure the repeatability of the test procedures.

An external hardware watchdog device is used to ensure a reboot of the system when some part of the software crashes. These watchdog devices are home built and very simple. The watchdog watches the serial port of the node computer for a ``hello'' string. This string has to be sent every minute (using a small program or even a script). If this string is not received within 90 seconds of the previous string, the watchdog circuit will cut the power to the setup and will turn it on after a minute. Using either journaling file systems or RAM based file systems minimizes problems caused by a fsck operation after the setup has been powercycled.

Figure 4: A typical WirelessLeiden Node setup.
\begin{figure}\begin{centering}
\psfig{file=node.eps,width=8cm}\small\itshape\end {centering}
\end{figure}

The software environment on the machines is a standard Open Source free UNIX (currently FreeBSD) distribution stripped down to fit in a minimal hardware configuration. The (kernel) device drivers that are used to control the wireless network cards and some network operations and management utilities are added (ref. section 2.10).

Using a standard off-the-shelf Open Source operating system enables us to implement a node quite fast while at the same time keeping the flexibility of changing things on all levels when the network grows and / or the technology changes. Another aspect is the large and diverse knowledge-base available within the development and engineering group. Also, the Open Source development model guarantees a fast turn around time in fixing bugs or evaluating features. Last but not least, in a not-for-profit organization working from donations the initial cost of the software is of major importance.


next up previous
Next: Hardware Platform Up: Method Previous: Site allocation
Rudi van Drunen 2003-04-08