The MPLS WG Archive

["James R. Leu" <jleu@mindspring.com>]

On Thu, Apr 27, 2000 at 10:01:37AM -0400, Curtis Villamizar wrote:
> 
> In message <4.2.2.20000427091619.00a8fd10@farley.cisco.com>, Jeremy Lawrence wr
> ites:
> > Jim,
> > 
> > At 05:13 04/27/2000 -0400, Jim Sullivan wrote:
> > >Jeremy, 
> > >
> > >
> > >At 06:47 PM 04/26/2000 +1000, Jeremy Lawrence wrote:
> > > >At 17:39 04/25/2000 -0700, Bora Akyol wrote:
> > 
> > [...]
> > > >>FEC to LSP binding is almost historical now.
> > > >
> > > >That's arguably true in router-based networks, because of the
> > > >predominance of TE as an application of MPLS in those networks.
> > > >However, FEC to LSP binding is very much alive in ATM MPLS
> > > >networks. 
> > >I'm not sure I completely follow this thread. I make the assumption
> > >that Bora believes all packets that do not get forwarded on 
> > >TE LSPs can be efficiently forwarded as unlabeled packets. 
> > >Is the point your raising that ATM-LSRs in general can not
> > >be expected to forward unlabeled traffic efficiently ? 
> > 
> > Exactly right. ATM-LSRs do not necessarily have any capability
> > of forwarding unlabelled packets. Most, but not all, ATM-LSRs 
> > have some capacity to forward unlabelled packets, but (in all
> > cases I'm aware of) this is limited compared to their capacity
> > for labelled packet forwarding.
> > 
> > So, where the 'default' forwarding mechanism in a router network
> > using MPLS TE can sometimes be ordinary IP packet forwarding (see
> > also the next paragraph), the default mechanism in an ATM MPLS
> > network normally must be hop-by-hop routed MPLS.
> > 
> > There are also several VPN schemes that require labelled packet
> > forwarding as the default in any type of MPLS network over which
> > they operate.
> > 
> > Regards,
> > 
> > Jeremy Lawrence
> 
> 
> Jeremy,
> 
> Some ISPs have considered running MPLS-RSVP backbones in which the
> interior nodes do run an IGP but do not run IBGP.  In such a network,
> any router which has one or more EBGP peers must run IBGP and have a
> tunnel to any other router with one or more EBGP peer.  The IGP
> carries only internal routes and is used for the MPLS signaling plus
> management of the network.
> 
> This design does not forward external IP traffic hop by hop and does
> not use LDP either.  A similar approach could be taken for MPLS-ATM.

I had a design built around an ATM network that did just this.  A real
plus to this idea is if the tunnels to the others EBGP speakers were
created dynamically as soon at they were learned. (ie each time a new
BGP next hop is learned a tunnel is created to that BGP next hop)

Is this a new idea or is this the same idea as BGP short cuts?

> 
> Curtis

-- 
James R. Leu