The total amount of BGP traffic is growning.
Nothing else has changed since the last NANOG.
From a small number of observations, it appears that making software problems and configuration errors (by engineers) are the major sources of routing instability.
There appears to be no correlation between announcements and widthdraws.
What is the source of this problem?
There are 10 ASes (< 1%) that appear in contribute 45% of the instability. Unfortunately, these top 10 also contribute 80% of the routes.
The top 8 are the same, but the last two vary. It seems that the instability is proportionally distributed by number of routes announced.
Of the top 1000 prefixes (~4%) causes about 5% of the instability.
Most instability occurrs during a period of less than 1 minute.
The period appears to be every 30 seconds. This looks like an IGP problem.
The averge AS path is 2.8. What does that mean?
There is a number of multiply announced prefixes. This seems to be growning fairly quickly.
Rob Libshultz says that he has observed that at least in his case, the stats reported on the servers did not seem to agree with what he observed on his router. Craig said that he has done period sanity checks on the data with peers that have logging turned one and he has no reason to suspect the data.
One NANOG participant postulated that the IGP periodicity must be attributed to RIP. Craig said that was probably reasonable.