Re: Just shy of OT MPLS Question

Next time I'll just email you directly. I thought it was an mpls thing. I
was told, (by a juniper employee no less) that there was no way to get
anything but IGP routes into INET.3 and even this was limited to ISIS,
OSPF and RSVP. I can't think of one good reason to distribute labels
using bgp since it would then become the IGP (bad) and then hope your
LSP/tunnel never has to cross the AS boundry (well at least on the
ciscos..). I supposed it's on me to look up the configuration. Good
catch!

Keegan

From:
Scott Morris <smorris_at_ine.com>
To:
ALL From_NJ <all.from.nj_at_gmail.com>
Cc:
Keegan.Holley_at_sungard.com, Cisco certification <ccielab_at_groupstudy.com>
Date:
12/21/2009 11:42 PM
Subject:
Re: Just shy of OT MPLS Question

Hmmmm... You're trying to make me think. :) Dude, it's the holiday
week, what's up with that?

Both vendors can exchange labels via bgp, but what you're getting down
with is really depending on how/what you are going to use to resolve
next hop reachability in the end. And therefore, how you are actually
going to pick your tunnel to use!

Juniper will use the inet.3 routing table whenever possible. But if you
look at their overall set of preferences (AD equivalent) you'll see that
"LDP" and "RSVP" are actually two listings there. So they take things a
little further than Cisco did with "autoroute announce". Just a way of
interpretation though. I can't think of anything significantly
different in LDP operations otherwise interop would suck beyond belief.

In general, LDP uses IGP because of the general (ok, it was ASSUMED)
behavior of link-state protocols. That and the convergence times of IGP
are generally faster than BGP timers as well, so overall reliability is
different.

Although you run BGP inside your own network, you run it for different
reasons than you run an IGP. Also, kinda hard to have a "bgp free core"
whilst running BGP. :)

But AFAIK there isn't anything in the spec that says one way or the
other. At least not that has leaped out at me in any reading I've
done. but it's entirely possible I've just glossed over it 'cause I've
never thought about it! heheheh... I think it's just per-vendor
implementations though.

*shrug* Perhaps after the holidays I'll remember this and be motivated
to lab it up. I have a few devices I can play with! (grin)

*Scott Morris*, CCIE/x4/ (R&S/ISP-Dial/Security/Service Provider) #4713,

JNCIE-M #153, JNCIS-ER, CISSP, et al.

JNCI-M, JNCI-ER

evil_at_ine.com

Internetwork Expert, Inc.

http://www.InternetworkExpert.com

Toll Free: 877-224-8987

Outside US: 775-826-4344

Knowledge is power.

Power corrupts.

Study hard and be Eeeeviiiil......

ALL From_NJ wrote:
> lol ... you got me. Yes, the LDP protocol exchanges labels based on ...
?
>
> I was not technical enough in my quick answer. ;-)
>
> I am not sure why Cisco chose this ... (other than lower overhead on
core /
> P routers)
>
> I am not very knowledgeable with Juniper, but I understand that Juni can
do
> this with BGP? Probably someone smarter than me knows more about Juni
... I
> just have not ever studied them.
>
> What saith thee Juni guys? Super Scott has some Juni certs, maybe we
can
> get the super Scott to respond.
>
> I am studying for a cvoice test! lol ... I think I am trying to fit too
> much into my little brain.
>
> Have a great night,
>
> Andrew
>
>
> On Mon, Dec 21, 2009 at 10:01 PM, <Keegan.Holley_at_sungard.com> wrote:
>
>
>> Thanks for responding. Technically the outer labels come from LDP or
RSVP.
>> They are then mapped to the route for the endpoints of the LSP/tunnel.
 I
>> was just wondering why this couldn't be built with BGP instead of an
IGP.
>>
>> ALL From_NJ <all.from.nj_at_gmail.com> wrote on 12/21/2009 09:43:40 PM:
>>
>>
>>
>>> Good evening Sir,
>>>
>>> I am pressed for time, so please forgive the quick and short
>>> response. Outer labels come from IGP, and not from BGP ... so no
>>> way to use BGP for this. HTH,
>>>
>>> Andrew Lee Lissitz
>>>
>>>
>>> On Mon, Dec 21, 2009 at 9:07 PM, <Keegan.Holley_at_sungard.com> wrote:
>>> Does anyone know why you cannot use BGP routes to build MPLS tunnels.
Is
>>> it the obvious (slow hello timers, possibly next hop ambiguity) or is
>>> there something inherent in the protocol suite (MPLS, RSVP, LDP) that
>>> makes it impossible.
>>> Not that I would ever want to... Just a little curious.
>>>
>>>
>>> Blogs and organic groups at http://www.ccie.net
>>>
>>>
Received on Tue Dec 22 2009 - 21:37:24 ART