The MPLS-OPS Archive

[Robert Raszuk <raszuk@cisco.com>]

Hi Umit,

Yes you are correct. In the above case R3 would change next hop to
itself only for MPLS-VPNs routes, but since this is not the case here at
all next hop would still remain the P1's interface address (for reg ebgp
session on the R3-P1).

Now this next hop has to be known to AS#1's IGP and label will be
alocated to it so "theoretically" core of AS#1 (R2 here) can be bgp
free. I said theoretically as if for example you run multicast at least
multicast-bgp AF has to be carried with all of those 10K routes which
are m-cast sources for RPF not to fail at R2.

Rgs,
R.

> Umit.Inan@alcatel.com.tr wrote:
> 
> Hello :
> 
> "So in the above topology, R3 may be learning 10,000 ipv4 routes from P1
> (which is an eBGP neighbor) and may advertise those 10,000 ipv4 routes to
> R1 and R4 (R3's iBGP neighbors). From R1 and R4 perspective, R3 is the
> next-hop for all those 10,000 routes"
> 
>  RFC 1771 & 2858 states that when a BGP speaker advertises the route to an
> internal peer, the
>  advertising speaker should not modify the next hop information
> associated with the route.
> 
> As far as I understand from this statement  for all routes learned from P1
> next-hop should be P1 instead of R3
> from R1 and R4 point of view,as long as next-hop self command is not
> practiced at R3 .
> 
> Please let me know  if I am wrong.
> 
> Best Regards ,
> 
> Rajiv Asati <rajiva@cisco.com> on 05/20/2002 08:01:47 AM
> 
> 
> 
>   To:          Christopher Poh <tvpoh@essex.ac.uk>
> 
>   cc:          mpls-ops@mplsrc.com(bcc: Umit INAN/TR/ALCATEL)
> 
> 
> 
>   Subject      Re: [MPLS-OPS]: AS Border Router
>   :
> 
> 
> Christopher,
> 
> In a MPLS network, an ASBR, which is learning ipv4 routes via an eBGP
> neighbor, typically doesn't allocate/advertise any label for those ipv4
> routes that are advertised to its iBGP neighbors.
> 
>          <-------AS#1-------->        <-----AS#2.--->
>          R1-------R2--------R3--------------P1---------------P2
>                  |
>                  R4
> 
> So in the above topology, R3 may be learning 10,000 ipv4 routes from P1
> (which is an eBGP neighbor) and may advertise those 10,000 ipv4 routes to
> R1 and R4 (R3's iBGP neighbors). From R1 and R4 perspective, R3 is the
> next-hop for all those 10,000 routes. So, R1 will have a label to reach R3
> and it will use that single label to reach all those 10,000 routes.
> Similarly, R4 will also have a label to reach R3 and that label will also
> be applied to any of those 10,000 prefixes.
> 
> As far as peering to all the internal routes are concerned, MPLS provides
> the ability to make core routers BGP free. In other words, R2 above doesn't
> need to be an iBGP neighbor and so it doesn't need to know about 10,000
> prefixes.
> 
> Please let me know if I could be of more help.
> 
> Rajiv
> 
> At 09:13 PM 5/19/2002, Christopher Poh wrote:
> >Hi,
> >
> >How does an AS border router advertise and make known to all other routers
> >within a domain about label/FEC binding information for every exterior
> >routes it
> >carry? Does the router need to peer with all the interior routers in order
> >to distribute this information?
> >
> >Hope someone could help me with this.
> >
> >
> >-------
> >The MPLS-OPS Mailing List
> >Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> >Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> 
> -------
> The MPLS-OPS Mailing List
> Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> Archive: http://www.mplsrc.com/mpls-ops_archive.shtml
> 
> -------
> The MPLS-OPS Mailing List
> Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
> Archive: http://www.mplsrc.com/mpls-ops_archive.shtml

-------
The MPLS-OPS Mailing List
Subscribe/Unsubscribe:  http://www.mplsrc.com/mplsops.shtml
Archive: http://www.mplsrc.com/mpls-ops_archive.shtml