[1343] in linux-net channel archive
Re: route propagation across administrative domains
daemon@ATHENA.MIT.EDU (Robert Sanders)
Thu Nov 9 00:58:23 1995
To: submit-linux-dev-net@ratatosk.yggdrasil.com
From: Robert Sanders <rsanders@mindspring.com>
Date: 07 Nov 1995 22:15:10 -0500
In-Reply-To: Raul Miller's message of 7 Nov 1995 17:52:49 +0100
On 7 Nov 1995 17:52:49 +0100, Raul Miller <rdm@tad.micro.umn.edu> said:
> This brings to mind the obvious question: why aren't routes "anchored"
> in place for five-tenminutes (or whatever) after they're initially
> established, to damp out this kind of thing? At least for where they
> affect admin domain route points...
cisco has released some beta code that implements BGP route flap
dampening. Once it's deployed on most of the NSP backbone routers,
that means that routes prone to flapping will be allowed to do so much
less frequently. However, simply applying a hard minimum to route
changes would adversely affect convergence time.
In general, some routing protocols have the concept of a "hold-down"
period, during which the existing route will be retained in the table
and new routes for that destination not accepted. RIP is such a
protocol; even after a route has timed out, a new route is not
accepted for several times the refresh interval.
-- Robert
--
MindSpring: use us and nobody gets hurt.
