RE: ( SP CCIE) MP BGP among PE's

From: Andrew Lissitz \(alissitz\) (alissitz@cisco.com)
Date: Sun Oct 30 2005 - 09:12:56 GMT-3

A hub and spoke would work, as long as the central router reflects the
routes from each spoke to the others. BGP slit horizon issues will
occur if the central router is not configured as a RR. If the central
router goes away ... there goes reachability as well.

I am not sure what you mean by the IGP question.

Here is a topology that I think you can run with only a few routers.
You only need a couple of smart routers (PEs) to make this work. Group,
please add or take away from this in order to make it better.

CE---PE1---<mpls>---P1---<mpls>---PE2
                               |
                               | (Un tagged link)
                              RR

Off of each PE you can hang a CE or create a loopback interface add this
loopback to a vrf. In either case you will want to have multiple VRFs
on each PE.

Keep the link between P1 and RR untagged for part of this and tagged for
other parts. It makes the show commands more interesting.

Make the link between CE and PE1 mpls. Use only BGP between the CE and
PE1 and see if mpls has trouble or not ...

You can do a lot with a topology like this. Group ... anything to add
to the topology?

________________________________

From: Jongsoo [mailto:bstrt2004@gmail.com]
Sent: Sunday, October 30, 2005 1:10 AM
To: Andrew Lissitz (alissitz); FORUM
Subject: Re: ( SP CCIE) MP BGP among PE's

I think I can also see a way to reduce the number of seesion by making a
hub-and-spoke topology.
Then I can reduce MP-ibgp session to 3 ( from one HUB PE to three spoke
PE), correct?
Of course, in this case, RIP or OSPF needs in VRFA in order to achieve a
full reachability between spoke?

Someone correct me if I am wrong.

Thanks

Jongsoo

On 10/29/05, Jongsoo <bstrt2004@gmail.com> wrote:

        Thanks Andrew and I got it.
        Basically, it works as a similar way as iBGP

        The workbook I purchased last wek and am studying now is very
poor.
        But I say ...better than nothing...I am even validating the
solution as well

        Jongsoo

        On 10/29/05, Andrew Lissitz (alissitz) < alissitz@cisco.com
<mailto:alissitz@cisco.com> > wrote:

                Good evening Jongsoo,

                Vpnv4 routes need to be propagated to each PE that has
this VRFA. On
                one of the PEs, you can do a show ip route vrf VRFA and
see the routes.
                This routing table needs to be converged, and thus you
need the routes
                from the other PEs. You will get these routes via
mpibgp sessions with
                other PEs or RRs.

                It is easier to use RRs and this way you are only
creating a session
                with the RR, and from the RR to the other PEs ... But of
course you
                already know this MR CCIE ;-)

                Does this answer your question Jongsoo?

                Andrew

                PS - Have you joined the comserv mail list? I joined
both ;-)

                -----Original Message-----
                From: nobody@groupstudy.com [mailto:
nobody@groupstudy.com <mailto:nobody@groupstudy.com> ] On Behalf Of
                Jongsoo
                Sent: Saturday, October 29, 2005 7:53 PM
                To: FORUM
                Subject: ( SP CCIE) MP BGP among PE's

                Let's say there are one VRF "A" and four PE's and each
of PE has one
                physical connection belonging to VRF A.
                The question is in order to make a full reachability of
VRF A networks
                among all four PE, the full MP bgp connection by
creating 6 MP-BGP
                session is a MUST?
                   Thanks
                Jongsoo

This archive was generated by hypermail 2.1.4 : Sun Nov 06 2005 - 22:00:55 GMT-3