Hi, On Tue, Nov 29, 2011 at 10:16:41PM +0800, Mark Tinka wrote: > On Tuesday, November 29, 2011 09:40:30 PM Gert Doering > wrote: > > > Most ISPs these days do not seem to have customers that > > use BGP-based multihoming using networks from the ISP's > > PA blocks - but of course, exceptions happen (we have > > one customer that uses an IPv6 /48 from our /32 due to > > historic reasons). > > We do have some customers that announce our own space back > to us because they multi-home to us and need some kind of > independent route control.
Well, yes. This is a scenario we have, of course, but I didn't
think of it because customer BGP sessions have completely different
- and very strict - route filters.
Customer session:
"customer may only announce a well-defined set of prefixes"
Peering or Upstream
"we accept most prefixes except for well-defined things we do not
want, like 224.0.0.0/6, subnets of our own address blocks, subnets
of IXP prefixes we're at, etc."
So, if a customer is multihomed *to us*, and uses Space from our aggregates,
this will be fine - different set of filters, and the routes are not
supposed to propagate anyway (when these filters are built, the "export
to world" community is not set, but "no-export" instead).
If a customer is multihomed to us and to some *other* ISP, and uses space
from our aggregate, we need to open the "Peering or Upstream" prefix filter
to take that into account. But that's not a typical case for us.
gert
--
USENET is *not* the non-clickable part of WWW!
//www.muc.de/~gert/
Gert Doering - Munich, Germany [email protected]
fax: +49-89-35655025 [email protected]
pgpg7KbZeAAHy.pgp
Description: PGP signature
_______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
