RE: WHy do LDP dont assign labels for BGP learned route ?

"Since the number of ASBRs in a network is much less than the number of BGP-learned prefixes, this practice saves a lot of state in the forwarding table, and saves a lot of control messages"

This is the reason why. In practical designs, the only labels that you need in the core of the MPLS network are for the Loopbacks of your PE routers (ASBRs). These routes are learned via IGP, not BGP. If the core is also carrying the Internet table via BGP, and LDP also generated labels for BGP routes, you would literally have hundreds of thousands of labels generated that you don't need.

If you want labels for your BGP routes you can generate labels via BGP+Label. To force this as the default design option via LDP would be bad practice. In my experience you're better off having the option to implement the feature if you choose, but not to be forced into it based on the default behavior of the protocol.

Brian McGahan, CCIE #8593 (R&S/SP/Security)
bmcgahan_at_INE.com

Internetwork Expert, Inc.
http://www.INE.com

-----Original Message-----
From: nobody_at_groupstudy.com [mailto:nobody_at_groupstudy.com] On Behalf Of Carlos G Mendioroz
Sent: Thursday, April 19, 2012 6:47 AM
To: CCIE KID
Cc: Cisco certification; CCIE OSL
Subject: Re: WHy do LDP dont assign labels for BGP learned route ?

Not sure why you call it a constraint, but it is an implementation fact.
In fact, it seems to be against RFC3031, section 4.1.2.2, that says that a router MUST bind labels to all the prefixes in its routing table.

I was puzzled by this too, and got someone to enlighten me, quoted w/o
permission:

Eric Rosen @ 01/06/2011 16:38 -0300 dixit:
>> The issue is that routers do not generate labels for BGP learnt >> prefixes.
>
> If one is using MPLS to send an IP packet from one BGP router (ASBR) > to another, across an ISIS or OSPF network, then within the ISIS/OSPF > network all one needs are labels for the IGP-learned prefixes.
>
> The early implementors of MPLS realized this, and decided to > implement it this way. You are correct that this is not precisely > what the spec says. I think this implementation practice is correct, > and I should have modified the spec to reflect it. I don't really > remember why I didn't, maybe the implementors didn't bother to tell > me about it until it was too late > > Since the number of ASBRs in a network is much less than the number of > BGP-learned prefixes, this practice saves a lot of state in the > forwarding table, and saves a lot of control messages.

-Carlos

CCIE KID @ 19/04/2012 06:23 -0300 dixit:
> Hi all
>
> I know that LDP wont assign label for BGP learned route. I just want
> to know the logic behind this idea. Is it a design constraint or else
> it is the implementation constraint ?
>
>
>

--
Carlos G Mendioroz  <tron_at_huapi.ba.ar>  LW7 EQI  Argentina
Blogs and organic groups at http://www.ccie.net