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

This paper presents the ETX metric for measuring paths in multi-hop wireless networks, which uses an expected transmission count rather than the standard minimum hop count. ETX thus takes into account loss rates and interference among different hops in the same path, leading to better routes that cause less interference, waste less energy, and get packets through more reliably. Although the metric is simple, it was interesting that it could work well and sometimes give a 2x increase in performance.

I have three criticisms for the paper. First, is the performance improvement really that good, or always necessary? Certainly 2x performance improvement is noticeable, but it only happens in a fraction of the cases in the evaluation. Is it worth switching to a more complicated routing metric? Second, ETX requires constant loss rate estimation, which could be costly and could also lead to misestimations, causing complex behavior. Minimum hop count is pretty robust and easy to understand, as long as you have a threshold for using reasonably good links. Finally, although the paper mentioned related work, such as the delivery-ratio-product metric of [33], and explains why it might perform poorly, this work is not evaluated in the experiments. How can we know whether ETX provides a 100% increase in performance over [33], or a 5% increase, or even a decrease? It would be more responsible to compare against these other metrics and see how much we really gain.

The paper was useful to read because ETX is probably a good metric to try out in any sensor network or MANET type of work, even if it turns out to be difficult to estimate correctly or to perform poorly in practice.