RE: Routing drafts

["Bernstein, Greg" <GregB@ciena.com>]

It's really a matter of how efficient one wants to be with optical and/or
TDM bandwidth.  
I'm not sure what you mean about carrying the information differently in
signaling and routing.  For SONET/SDH signaling we specify the number of
STS-1s, VC-4s etc that we want with simple relatively small integers.  Not
with floating point numbers (or collections of floating point numbers).
Hence we already made this decision with signaling and would like routing
consistent with it.

I'd like to update bandwidth often (from the optical networking perspective)
and keep the impact to the control plane to a minimum (since it doesn't take
much info to convey these bandwidth updates).  

Greg B.
-----Original Message-----
From: Kireeti Kompella [mailto:kireeti@juniper.net] 
Sent: Tuesday, February 26, 2002 9:52 PM
To: Bernstein, Greg
Cc: ccamp@ops.ietf.org
Subject: RE: Routing drafts


Hi Greg,
Snip...
> (b) Big issue -- The parameters for representing bandwidth on a link are
not
> very appropriate for TDM signals or WDM signals.  I've included below some
> more explanation taken from the IPO working group draft. However, this is
> the same thing that led to us breaking out the traffic descriptor stuff in
> GMPLS signaling for the SONET/SDH case.  This is really needed here too.

I don't know that this is _really needed_.  Note that the bandwidth
encoding in signaling is a floating point number.  Perhaps this isn't
the most aesthetic way of carrying TDM signal information, but it works
adequately.  Is it worth carrying the information differently in
signaling and in routing, and differently for each switching type?
What is the tangible gain?

Which is not to say we can't discuss this.

Kireeti.