Roof Network

John Bicket, Daniel Aguayo, Sanjit Biswas, Robert Morris, "Architecture and Evaluation of an Unplanned 802.11b Mesh Network"

I really liked that hand-on project. Ashima has written a nice summary for it on her blog.
I will only write about a couple of details I found interesting about the mechanisms:

• standard 802.11b hardware with RTS/CTS disabled, really? What is the benefit of disabling RTS/CTS?
• single-channel
• self-configuring network on preinstalled Linux boxes
• network is a mesh with many routing choices
• routing protocol: each node maintains partial matrix of link metrics between pairs of nodes, each node uses Dijkstra's algorithm for source-routing (not scalable)
• link metrics communicated through 1. packet header when forwarding, 2. flood queries when route to destination is missing, 3. nodes overhearing responses to flood queries
• probing of link quality through periodic broadcasts
• link metric given by adding inverses of throughput
• overriding default selection of transmit bit-rates of wireless hardware for optimization

I found interesting about the evaluation the detailed figures on e.g. the relation between throughput and distance, especially the figures on how the network performance drops when nodes are dropped.

I was missing a discussion on packet collisions which is relevant because of single-channel and omni-directional antennas.