I’m not sure on what use this data, if collected, would be. Latency is the most important. It's not operationally useful in any way that I can think of, but it is interesting (at least to me). It's possible that Bill has something in mind, though.
[...] which is more meaningful a metric, AS-path or hop count? Good point. Before we can decide which is more useful, we have to decide what they would be useful for. But, I think it's not really feasible to analyze hop count, because you would have to collect that data with a huge number of traceroutes. Average/median AS-path length can be estimated by static analysis of BGP tables from various routers.
Many networks have a large number of routers but the packets don’t stay in them very long. It wouldn't be a very good router if the packets hung around for a long time before leaving :)
On Wed, Nov 21, 2018 at 11:26 PM Ben Cannon <ben@6by7.net> wrote:
This begs the question, which is more meaningful a metric, AS-path or hop count? Many networks have a large number of routers but the packets don’t stay in them very long.
-Ben.
On Nov 21, 2018, at 8:10 PM, Bryce Wilson <bryce@thenetworknerds.ca> wrote:
I don’t have any hard statistics but I notice that on a majority of ASs on bgp.he.net, the average AS path length is between 4 and 5. As for the average number of hops, it clearly depends on what type of traffic and many ASNs have more than one router. Going on my own experience I would say between 8 and 10 hops would be the average of non-cached content. If you included cached content such as cdns and caches then the actual average might be closer to 5 to 7. This is only an estimate from my own network and those of my clients so the actual value may be completely different.
As with what others have said, I’m not sure on what use this data, if collected, would be. Latency is the most important.
Thanks ~ Bryce Wilson, AS202313, EVIX, AS137933