RE: Fwd: Comments on GMPLS signalling drafts

["Vishal Sharma" <vsharma87@yahoo.com>]

Manoj,

I just got to my email after a hiatus, and saw your question on
tunnelling at the end of this email (I've deleted all the rest).
Since there seems to have been much debate on this, and you
brought it up yet again, I thought it appropriate to answer.

Actually, I don't understand where the confusion is. I think John Drake
already answered this question, both in his reply directly to you,
and again, in his observations on Juergen's email (the one where
Juergen tried to explain his understanding of FA's etc.).

The basic answer is that the lambda LSP will be advertised in routing,
and stored in the TED at A and D, as a TDM-capable FA LSP.
From that point forward, nodes A and D will exchange TDM labels for
any TDM LSP being setup over this FA LSP(those labels will of course
follow all the layering rules that Maarten and Juergen have been
talking about, and the label requests themselves will be tunnelled
directly from A to D; for example, over the SONET/SDH section or line
DCC). In fact, one could view this FA LSP as a "one-hop"
TDM link, with all the properties of a TDM link.

More specifically, all TDM LSPs tunneled through the
FA LSP will *not* be allocated the same label (this is a physical
impossibility in a circuit switched network in any case). Rather,
they will each be allocated a label that denotes the time slots
on the FA LSP (now viewed as TDM capable) that each TDM LSP will
use.

(BTW, you might also find a presentation
I gave at MPLSCon'01 in March this year on this very subject useful to
clarify some of your doubts (there are some pictures there that
are useful for visualizing these scenarios).)

Finally, I saw in one of your earlier emails on this subject a request,
which seemed to boil down to asking that every possible scenario
be outlined in some document or the other. While, in general, I am
supportive of making things clear (and I've expressed that viewpoint
several times on this list and the MPLS list over the past couple years),
in this particular case, I think you might be asking for too much!
Doing that is practically impossible.

Once the building blocks are specified (and I support that they be
specified as clearly as possible), the compositions of these blocks
is something that has to be done depending on the specific applications
that one intends to use these blocks for.

-Vishal

-----Original Message-----
From: owner-ccamp@ops.ietf.org [mailto:owner-ccamp@ops.ietf.org]On
Behalf Of manoj juneja
Sent: Friday, December 14, 2001 3:33 PM
To: lberger@movaz.com
Cc: Eric.Mannie@ebone.com; dimitri.papadimitriou@alcatel.be;
ccamp@ops.ietf.org
Subject: Re: Fwd: Comments on GMPLS signalling drafts


Hi Lou/Eric/Dimitri,

  Please see the following para from gmpls-architecture draft :

<snip>

>From: Lou Berger <lberger@movaz.com>
>To: "manoj juneja" <manojkumarjuneja@hotmail.com>
>CC: <Eric.Mannie@ebone.com>,<dimitri.papadimitriou@alcatel.be>,
><ccamp@ops.ietf.org>
>Subject: Re: Fwd: Comments on GMPLS signalling drafts
>Date: Fri, 14 Dec 2001 15:18:34 -0500
>
>see comments inline.
>
>At 09:10 PM 12/13/2001, manoj juneja wrote:
>
>
>
>> >From: "manoj juneja" <manojkumarjuneja@hotmail.com>
>> >To: ccamp@ops.ietf.org
>> >Subject: Comments on GMPLS signalling drafts
>> >Date: Tue, 11 Dec 2001 16:15:10 -0700
>> >
>> >Hi All,
>> >
>> >1. SE-style supported by GMPLS or not ?
>> >

<<snip>>

>> >8. The LSP hierarchy concept is still not clear. Some days back I posted
>> >one
>> >doubt related to tunneling of TDM LSP over Lambda LSP using the concept
>>of
>> >forwarding adjacency and different people replied with different
>>thoughts.
>> >Does this mean this concept is not standardized in GMPLS ?
>> >My question was :
>> >
>> >"If there are 4 nodes say A, B, C and D. There is a Lambda FA
>> >established from A to D and if a new TDM LSP request comes to node A
>> >which is to be tunneled through the already established lambda FA-LSP
>> >then the node A sends the Path/label request message directly to node
>> >D. What label the node D will send back to node A in the RESV/label
>> >mapping message since the FA-LSP is just one label (lambda) ? Does it
>> >mean that all the LSPs which are tunneled through the lambda FA-LSP
>> >will be allocated the same label by node D to node A ? If this type of
>> >scenario can't exist in GMPLS then please let me know that too."
>
>I believe this topic is being covered in other threads.
>