Network Topology of ISI Wireless Sensor Network Testbed

---

Among 30 tesbed nodes, 14 nodes are permanently placed in offices of ISI. The following graph show the node distribution on 10th and 11th floor. Despite of careful node placement, it appears that toplology of the network changes all the time, mainly due to unpredictable charateristics of radio propogation for in-door environment. Asymetric, highly lossy or osillated links appear common, which in turn confirms that the real world is always the only way to validate our design and the driving force for new research problems.
We have been developing tools to aid our expreiments. Particularly, we use a tool called linkScan to roughly map the topology of the network, which is helpful to design experimenet scenarios.

Topology Mapping by linkScan

LinkScan is built upon directed diffusion API version 3. The source code is avaible in scadds cvs repository. The tool consists of two components:

• linkScan: A diffusion filter application: This module overheard all the packets from radio and construct the neighbour list. Items in neighbourlist is cached and will be timed-out if not get refreshed periodically. linkScan also provides publication service of linkScan using directed diffusion API. 
• linkApp: A diffusion application: This program expresses interest in linkScan data by subscription interface in diffusion API. It collects neighbor lists from nodes and print it out.

To start LinkScan, you may place the following in /usr/bin/usrapp.sh

diffusion &    # start main diffusion module
gradient   &    # start main diffsuion routing module using gradient reenforcement
linkScan  &    # start linkScan module

Notice that linkScan module can be started simutanously with any other diffusion applications. Without linkApp running, it only overhears the traffic and never generates any packet to radio interface thus does not interference with other diffusion applications.

After reboot nodes, you can start linkApp at one node to collect links throughout the network. for example

[zhaoy@scadds zhaoy]$ rsh104 18 linkApp
Diffusion Routing Agent initializing... Agent Id = 1188
Neighbors of Node 31: 22 37 36 18 26 29.
Neighbors of Node 22: 36 18 31 26.
Neighbors of Node 37: 29 36 31.
Neighbors of Node 18: 22 26 36.
.......

You can also start linkApp using interactive mode to print neighbour list for particular node.

[zhaoy@scadds zhaoy]$ rsh104 18 linkApp -i
Diffusion Routing Agent initializing... Agent Id = 1193
Link scan is running
n) collect neighbour list for a particular node
a) collect neighbour list for all nodes
t) toggle beeping switch
e) exit.
n 
Enter the node ID:36
Neighbors of Node 36: 37 22 31 18 26.
Neighbors of Node 36: 37 22 31 26.
.....

The interactive mode is particularly useful when you place testbed nodes incrementally. With the feedback from linkScan, you can add nodes one by one form the basic topology backbone and then add addtional nodes to create interesting scenarios.

Visualization of linkScan

We also write a script map104 to visualize the data from linkApp. This script periodically generate PNG graphs. For example,
the following graph is available at  the topology mapping for our testbed in real-time. (Not available when  other experiments are running. :-)

TO-DO LIST

LinkScan is quite preliminary as a mapping tool. We can extend linkScan to estimate the loss rate of links by overhearing beaconing messages with sequence #. Additionally, we can also extend linkScan to collect other exprimental data such as some statistics of performance metrics.

---

Last modified by Jerry Zhao : 1/14/2002