21. A High-Throughput Path Metric for Multi-Hop Wireless Routing

This paper is trying to solve the connection problems on multi-hop wireless networks like the previous paper on Rooftop network. But instead of having the antennae on the rooftops, this time it is placed in a university building with 29 nodes totally. The authors didn't really mention how they chose the locations of their hotspots, however. And they put it clear that mobility of these wireless access points are not put into consideration.

The factor they are trying to change in the 802.11b network this time, is the hop counter for routing protocols. It uses an Expected Transmission Count (ETX) which incorporates the ideas of:

• masking transmission errors, so that all but the worst connections appear loss-free
• finding a path with the fewest expected number of (re)transmissions
• predicting packet loss ratios in both directions of each link, so asymmetrical link speeds are taken into consideration
• penalizing interference between hops on multi-hop network

ETX = 1 / ( forward delivery ratio * reverse radio ), assumptions:

1. for protocols with link-layer retransmissions, such as 802.11b
2. fixed transmit power for the wireless

The fact that ETX does not avoid congested links makes it free from load-adaptive link delays. It also prefers paths less than 3 hops even when 4+ hops have better throughput, and avoiding wireless interference.