At 9:15 AM + 6/27/03, Mwalie W wrote:
Hi All,
What causes Circuit Flapping? While checking the Network Notices of a
certain ISP, I see this as a common problem.
This problem can not be solved for all time?
Thanks.
Mwalie
I'm assuming you are referring to BGP flaps in the global Internet.
The simple answer is no, it can't be solved unless you never add or
remove a router or AS, bring a link up or down, etc.
You can, however, minimize it, but there are tradeoffs in every way
you do it. The fundamental tradeoff is the more stable you make the
network, the longer it will take to converge. Now, converge is a
relative concept when it comes to the global Internet. I doubt you
would find any routing researcher who believes the Internet ever is
in a stable, converged state. There's too much change.
One of the best ways of reducing flaps from links going up and down
is for AS to advertise only aggregate routes that resolve to the null
interface. Assuming they have a sane, aggregatable address structure,
that will prevent the rest of the Internet from learning about link
failures and many routing failures. The increased stability comes at
the cost of blackholing, but if destinations blackhole, you'll still
find out about it from such things as TCP timeouts.
The overall question of Internet stability remains a complex research
subject, and I'll have to admit it's probably getting worse. I am
involved in the Internet Research Task Force (an IETF subgroup)
Routing Research group, and there certainly is no consensus even to
the direction of solution. The current model of BGP and addressing
is in trouble, not from address exhaustion but route churn and
instability.
Flap is only one case of instability. I discuss some of them in my
book, Building Service Provider Networks, and also in some
CertificationZone BGP tutorials. You can browse the NANOG and RIPE
sites for presentations, and also Vern Paxson's PhD Dissertation,
End-to-End Packet Dynamics in the Internet, is an important
reference available online. Also, search for presentations on global
convergence by Craig Labovits and the late Abha Ahuja.
There really are three cases of Internet-related stability: global,
AS, and single router. For some work on single router, see our draft
in the IETF Benchmarking Methodology Working Group,
http://www.ietf.org/internet-drafts/draft-ietf-bmwg-conterm-04.txt.
This is in the final process of editing before becoming an RFC. Our
methodology draft has expired, but I'll put a copy online soon and
hope that the team gets back to work on it -- this is not a good
economy for research!
Again, I've only skimmed this topic.
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7i=71572t=71518
--
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]