BGP update propogation times


     Last time I had any reasonable idea as to the time it took for
network updates to propogate across the various backbones, I concluded
that it took approximately 15 minutes for reachability via BGP to be
advertised to what seemed like a widely dispersed set of locations.
Some of them were connected either directly or indirectly to the MCI
backbone, but others were connected to ANS or Sprint. Today, we had a
scheduled outage of our main routing core here at Penn. We powered all
of the routers back on at approximately 09:00 EDT.

     Until approximately 10:20 EDT, we were unable to perform
nameservice lookups due to the root name servers not sending responses
to us. Between approximately 09:30 EDT and 10:20 EDT, we were able to
get to more and more of the root nameservers, but it wasn't untill after
10:00 EDT that any of the root namerservers were reachable. Does it
really take over 30 minutes for routes to propogate across the major US
backbones? [I'm ignoring the fact that we were unable to get answers to
name service queries from the root name servers until after 10:20 EDT -
I don't know what caused it, but for the purpose of determining how long
it takes for reachability to propogate, that fact that I could ping or
traceroute to the servers means that my nets were reachable.]

     Has anyone done any testing on this to try to determine the
propogation time? Was there some major problem this morning on MCI or
Sprint or some NAP? I have a net that I could use for this testing, but
I need a good selection of hosts to try to ping [preferably at least one
on each network providers network - ie. one for MCI backbone, one for
JVNCnet/GES, one for Sprintlink, one for bob's stop-n-shop dial-up
service, etc.]

     By the way, is there still a large amount of route flapping going
on in the core routers?