The MPLS-OPS Archive

[Christopher Lewis <chrlewis@cisco.com>]

Yakov,

So my current understanding is as follows:

In any MPLS network you need LDP. In most large provider IP networks you 
tend to have BGP deployed, so both of them tend to be there to begin with 
for the providers considering deploying these L2 services.

I don't think there are many arguments against using BGP for auto-discovery 
as its nature is to distribute all information to all peers. There is 
of  course the DNS option for those that like that thinking. That leaves 
the debate over the use of BGP for signaling. In the case of L2 services 
like frame and ATM, it is multiple point to point connections that must be 
made. Given that is the case, a directed LDP session to exchange 
information just between the two end points connecting seems more efficient 
than using BGP to tell all nodes about something that only 2 of them care 
about. I can see contrasting arguments for the case of multipoint ethernet 
connections, but from what I see currently, those are not a significant 
portion of the market opportunities, the majority of ethernet connections 
planned over MPLS are still point to point.

As is often said, the market will decide.

Chris


At 05:04 PM 11/16/2002, Yakov Rekhter wrote:

>If one uses BGP for autodiscovery, then I've yet to see any convincing
>*technical* arguments why LDP for signaling is any better than BGP.
>In other words, why BGP for autodiscovery + LDP for signaling
>would be any better than BGP for both autodiscovery and signaling.

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