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

This paper describes the underlying routing protocol used in Roofnet, called ETX (expected transmission count). The key idea is that wireless medium breaks the assumption that link is either working or not working. Minimal hop count can be suboptimal because the smaller the number of hop is, the chance of a weak link increases. Instead, ETX considers three factors: transmission loss ratio, assymmetry in the network link quality, interference causing reduction in throughput as the number of hops increase.

I liked the paper because this scheme neatly integrats all of the three mentioned dimensions that affect the throughput of the route. It also has extensive evaluation to prove the ETX is more optimal than minimum hop count in most cases.

The environment described in this paper is very similar to that of Roofnet. Certainly in a loosely connected Roofnet environment, because ETX considers link quality in its routing decision, it provides better quality routes. However, if the environment is changed to a densely connected, corporate environment, would ETX's overhead of link probing make it unsuitable candidate to use in these settings? Even in roofnet, some parts of it (i.e. near MIT campus) would be much densely connected. In these settings, I would argue interference is much more important factor to consider than loss ratio. Smart channel allocation will probably provide a much better gain in throughput.