Hi Anton, I suppose its to be expected that smaller blocks will flap more than larger ones if you consider that if I have a /8 I'm likely injecting the /8 into BGP from a lot of core routers and so its unlikely that I'll have a problem which takes out enough routers for my route to withdraw, by contrast I'd expect a /24 to be sourced from probably a single point and hence be affected by any issue at that particular PoP, just my 2-euros :) Steve On Tue, 27 Jan 2004, Anton L. Kapela wrote:
...They seem to come & go as they please!
Inspired by RAS's prior posts regarding /16's that were hastly de-agged, I took some time this weekend to answer a few questions which came up; how many updates do I hear about a specific prefix, and does the length have any relationship? Well, several hours later, I knew pretty straight away how many updates I heard about various prefixes in the STS Telecom looking glass, at least.
Consider this v0.0.1-alpa of the "Most-Heard Prefix" list, a sort of web-based and rule-less implementation of 'sh ip bgp flap-statistics.'
The sorted list, updated every 60 minutes, is available here in gzip format due to excessive size (indeed, 10 peers act as a amplifier):
http://eng.ststelecom.com/bgp-data/top-prefixes.txt.gz
Incidently, many of the most-updated prefixes are included in the list of RAS's /16 de-aggs.
Todo: ignore the effects of a peer session resets (i.e. don't consider a reload as part of the per-prefix flap count) and look at path changes with each update (is it a normal flap, or mid-day traffic-engineering session gone wrong?).
Any comments or suggestions for changes are very welcome; please reply off-list.
Thanks,
--Tk