Dampening considered harmful? (Was: Re: verizon.net and other email grief)
Jerry Pasker
info at n-connect.net
Mon Dec 20 22:51:42 UTC 2004
>An even more important consideration is whether our current paradigm
>of flap dampening actually is the behavior that we want to penalize.
>
>If a single link bounces just once, then thanks to our mesh,
>confederations, differing MRAI's etc., we can see many many changes
>to the AS path, resulting in dampening. Do we really want to
>inflict pain over such an incident?
>
>Tony
>
Of course not. Dampening should be set on a router according to that
router's views. The more chances a router can see a single event
multiplied, the more lax dampening should be on that router.
I think Rob Thomas's saying of "know your network" really applies here.
In my book, the threat of dampening to anyone not playing nice is the
true value of route dampening. Automatic enforcement of etiquette.
-Jerry
.
More information about the NANOG
mailing list