LDP or RSVP (no TE scenarios)

[pzrfxr]

If I wanted to simply build LSPs (assume no TE nor any plans for TE in the future), which signalling protocol would be the best to use (assume vendor neutral)? Would appreciate a rational as to why you would chose one over the other.

[Harold Ritter]

By mentioning that you are not interested in TE, the choice to me seems obvious that you want to use LDP.

RSVP is used for signalling traffic engineering tunnels and building a full mesh of LSPs between PEs would be a pain with TE (RSVP). Much easier with LDP.

Hope this helps,

[pkhatri]

Howdy,

In some of the networks I have worked on, we've tended to build a full-mesh of RSVp tunnels between the core devices and then run LDP from the edge devices to the core. That gives us the ability to do TE within the core, if required at any point, without incurring a large number of RSVP tunnels which we would have if we extended RSVP to the edge.

Paresh

[Harold Ritter]

Paresh,

I have also worked with many customers who have implemented this approach. It is just that the original poster specifically mentioned that he wasn't interested in TE.

Hope this helps,

[pkhatri]

My apologies, Harold.. I wasn't insinuating that you hadn't. I was just saying that it helps to go with a design like that in case the customer changes their mind in the future and wishes to do TE in the future.

Regards,

Paresh

[Harold Ritter]

Don't worry Paresh. I didn't take it as such.

You are right that the approach you suggested is way more scalable than trying to mesh all of the PEs in case you want to use TE.

The only point I wanted to make is why deploy TE right now if you have no intention of using it at this point in time. Implementing TE in the core later on will not necesseraly be more difficult.

Hope this helps,

[pkhatri]

Thanks for clarifying that Harold. I have way too much respect for you to be questioning any of your suggestions !

Anyways, the original poster has a couple of different options he can consider...

Cheers,

Paresh

[Harold Ritter]

It's always a pleasure for me to have good discusions with you and the other contributors on this forum.

Thanks,

[pzrfxr]

Harold and Paresh, you both provided good insight. Yes, I wanted to keep it simple to establish a level playing field in suggesting the best label distribution protocol for simply building LSPs -- without considering TE.

Thanks you both.

Bob

[olorunloba]

It does sound like a good design. But the applicability would depend much on the network topology.