BGP Confederations are a good solution for some problems along this
line, though I don't now how will that ties in with all the MPLS
world.
IGP (EIGRP, OSPF, ISIS, whatever) so you can get from router to router.
iBGP for clusters of routers that tightly connect on a LAN or redundant
links, whether interface->inteface or loopback based
iBGP route reflectors if there more than a half-dozen routers in each
cluster.
eBGP confederation member to confederation member (typically city<->city)
eBGP to talk to the world...
Strangely enough for Cisco, this still usally works reliably, whilst
scaling madly, the bigest problem is parts of the problem reuireing
more than the 128M limit on older routers...
George
> From cisco-nsp-request@puck.nether.net Mon Feb 5 09:10:19 2001
> Resent-Date: Mon, 5 Feb 2001 09:10:05 -0500
> Received-Date: Mon, 5 Feb 2001 09:05:33 -0500
> Date: Mon, 5 Feb 2001 09:05:17 -0500
> From: Eric Osborne <eosborne@cisco.com>
> To: Vinod Anthony Joseph Cherunni <vac@antarix.net>
> Cc: cisco-nsp@puck.nether.net, Philip Smith <pfs@cisco.com>
> Subject: Re: [nsp] BGP Advertisements
> References: <OF845E86B1.1DEADAB2-ON652569EA.004BC325@dsqworld.com>
> Content-Disposition: inline
> User-Agent: Mutt/1.2i
> In-Reply-To: <OF845E86B1.1DEADAB2-ON652569EA.004BC325@dsqworld.com>; from vac@antarix.net on Mon, Feb 05, 2001 at 07:24:43PM +0530
> X-GPG-Fingerprint: 6412 0836 E440 B3EA 980C 4951 611E 1819 2E71 8562
> Resent-From: cisco-nsp@puck.nether.net
> X-Mailing-List: <cisco-nsp@puck.nether.net> archive/latest/5072
> X-Loop: cisco-nsp@puck.nether.net
> Precedence: list
> Resent-Sender: cisco-nsp-request@puck.nether.net
>
> On Mon, Feb 05, 2001 at 07:24:43PM +0530, Vinod Anthony Joseph Cherunni wrote:
> > Dear All,
> >
> > Thanks once again for the advice. Just a few questions.
> >
> > As far as unnumbered, my .02 is that it's not worth the hassle. Using
> > unnumbered to customers means you need to have enough address space in
> > your loopback so that you don't have two customers with the same IP
> > address connected to the same box. And IP unnumbered in the core your
> > network is a Bad Idea; *very* difficult to troubleshoot problems if
> > traceroute doesn't tell you anything about which physical link you
> > cross.
> >
> > Question: If I am looking at using IP Unnumbered for NON-MPLS circuits,
> > Could'nt I do good with one loopback address per aggregation router?
>
> ...but then traceroute through your network doesn't show the links you
> cross, only the RIDs. up to you as to whether this is good or bad...
>
> >
> > Question: Is it preffered to use IBGP on aggregation routers, & announce
> > customer prefixes into the network using IBGP, instead of treating them as
> > external / stub links ie in the case of OSPF reducing overhead in the IGP,
> > & also reducing the floating of LSA's.
> >
>
> I thought this is what Phil and I just got through
> un-recommending...:)
>
>
>
> eric
>
> > Kindly advice me.
> >
> > With warm regards,
> > Vinod.
>
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:27 EDT