characterizing BGP updates (research topic?)

Sun Mar 27 21:47:38 UTC 2011

Under a different subject Bill Woodcock <woody at pch.net> wrote:
>
> On Mar 25, 2011, at 10:51 AM, Patrick W. Gilmore wrote:
>> The question is whether "some data" is better than "no data".  Honestly, I'm not sure.
>
> Yes, Patrick, I was just trying to be diplomatic about saying "not such a good idea" so he'd keep reading through to the end, where I suggested some other, possibly better, research topics.  I am, however, certain that other people could suggest even better research topics.  Good research topics are in demonstrably short supply among networking grad-students, so if y'all want to be helpful...

I'm currently looking for data characterizing BGP updates.
This page calls out 'bad' guys:

http://bgpupdates.potaroo.net/instability/bgpupd.html

but I haven't found anything yet which defines the terms or methodologies
used.  Is there a key for this page somewhere?  Are there similar
data sources for BGP update stats that go into more detail?

I'm personally interested in being able to tell the difference between
new announcements/withdrawals of a prefix to the Internet as a whole
(insufficient redundancy?) vs. changes in best path (next hop).  If I can get
my hands on raw BGP update traffic info, I would even consider
taking a stab at doing the analysis myself.

Pointers to data/previous results (as well as comments on why
this is a stupid question) are welcome.

Thanks,
Bill Bogstad