[zbisch]

I'm actually doing some research on this topic right now in grad school. I think there is definitely a lot more that can be done in this area; we're looking at things like time-of-day congestion, traffic management policy, actual application performance, and reliability. I think, as you said, there is a lot more to this than just bandwidth measurements. For example, last November a bunch of DNS servers went down in Comcast's networks that essentially meant no Internet for the majority of people that don't know how to configure public DNS.

We're working to eventually get a much finer grain survey of performance (e.g. a neighborhood), by piggy backing on network-intensive applications (currently BitTorrent). In addition to throughput, we're looking at things such as packet loss, latency (the last-mile in particular), time-of-day effects, as well as using traceroutes for underlying forward-path analysis.

My point is, if anyone has any ideas for other applications (something like Netflix would be great) to leverage, metrics to include, or advice, I'd be happy to hear them. :)