[c-nsp] BGP Fast-external-fallover
David Barak
thegameiam at yahoo.com
Sun Mar 13 10:43:33 EST 2005
--- Daniel Roesen <dr at cluenet.de> wrote:
> On Sat, Mar 12, 2005 at 09:43:59PM -0800, David
> Barak wrote:
> > Wait, your provider is dampening you, their
> customer?
> > Something's not right here - generally dampening
> is
> > used toward peers, not customers, because
> dampening is
> > the converse of the SLA approach. Personally, I
> > wouldn't pay for service from a provider which
> would
> > willfully permit me to remain down...
>
> Actually, your provider does a service to you. When
> you fixed your
> flapping for good, you can call your provider and
> have him reset
> the damping holddown.
So the burden goes from the provider to the customer?
As Randy Bush used to say, I encourage my competitors
to do this.
> If he would pass this flapping
> on, all/most/many
> peers/upstreams of him would have your prefix
> dampened, and almost
> nothing he can do about it. You stay down until all
> their reuse
> timers expire.
>
But if you postulate an enterprise with N locations,
the other N-1 locations will find the location with
the flapping circuit completely unreachable, rather
than only partially unreachable.
> Propagating the instability hurts more than damping
> it early at the
> source.
Dampening, like bogon filtering, is something which is
a good idea in theory, but in practice, can do far
more harm than good, unless it's very, very carefully
managed.
David Barak
Need Geek Rock? Try The Franchise:
http://www.listentothefranchise.com
NEW ALBUM, "The Sound and the Furry" available at
http://www.cdbaby.com/thefranchise
__________________________________
Do you Yahoo!?
Yahoo! Mail - 250MB free storage. Do more. Manage less.
http://info.mail.yahoo.com/mail_250
More information about the cisco-nsp
mailing list