A few people have mentioned it to me already, so...
TOP 20 Unstable Origin AS Rank ASN Upds % Upds/Pfx AS-Name [...] 3 - AS786 72205 3.6% 353.9 -- JANET The JNT Association
That hits a little close to home for a routing working group co-chair! My defence is that I do not believe it was our fault (and I'm certain it wasn't intentional on the part of our peer either). A peer of ours close to Geoff's monitoring point appeared to flap a number of its peers across the LINX's Extreme LAN. This is coincident with a renumbering that is happening on that LAN, but I've not been in touch with the peer ASN yet to see if it was connected, the log messages suggest it may have been. If you look at the second table on Geoff's page, ordered by 'updates per prefix': <http://bgpupdates.potaroo.net/instability/bgpupd.html> AS786 is clustered with a few other ASs at about 450 updates per prefix, but as we advertise 200 prefixes, that means we generated about 90,000 updates compared to 450 for peers advertising one prefix or 900 for two prefixes. Compare the pattern of updates for one of our prefixes, 128.86.0.0/16: <http://bgpupdates.potaroo.net/cgi-bin/pfx-rate-plot?prefix=128%2e86%2e0%2e0%2f16> With, for example, one of the BBC's prefixes, 132.185.0.0/16: <http://bgpupdates.potaroo.net/cgi-bin/pfx-rate-plot?prefix=132%2e185%2e0%2e0%2f16> I'll follow up with the peer next week to see what they did, I am going to guess at fat fingers, but I suppose the moral, as always, is what you see depends on where you're looking from. :-) <Cue the usual plea from people that measure this stuff for more probe locations, BGP peerings to monitoring points, etc, etc.> Cheers (defensively), Rob