RE: GMPLS dynamics vs. issues with optical amplifiers

["Marco Ruffini" <ruffinim@tcd.ie>]

Hi,
My understanding is that supposing that the optical path can be physically
allocated (e.g. none of the optical trails included in the path get blocked
for physical constraint reasons), there must be an adaptation in the
operative parameters of the optical amplifiers to avoid that the setup of a
new wavelength could disturb the signals brought by the other wavelengths in
the same fiber.


So on one hand GMPLS should consider if there is any blocking physical
constraint that may impair the allocating of a new wavelength on a specified
path.
On the other hand it should be considered how long it would take to the
transport layer to adapt all the physical parameter in the path to include a
new wavelength on the fiber without disturbing the existing signals (I
imagine this could be quite a lengthy process if more amplifiers are
cascaded, like in a long haul system for example). This issue I think would
put a lower limit to the reconfiguration-time of new optical paths.  

Marco

-----Original Message-----
From: Kireeti Kompella [mailto:kireeti@juniper.net] 
Sent: 10 April 2006 16:31
To: Marco Ruffini
Cc: ccamp@ops.ietf.org
Subject: Re: GMPLS dynamics vs. issues with optical amplifiers 

Hi Marco,

On Mon, 10 Apr 2006, Marco Ruffini wrote:

> I'm a PhD student on optical network and I'm following the development of
> the GMPLS work.
>
> I had a recent discussion with people working on the optical transport
> network who where telling me their concern about creating dynamical path
> over trails including optical amplifiers.
>
> They say that cross-talk and gain fluctuation effects may limit the
> capability of fast allocating new wavelengths over those trials.

We've talked on and off about these in CCAMP, and there have been 
drafts to describe some of the various optical impairments to consider.

In the GMPLS structure defined so far, there are link attributes (such 
as metrics and bandwidth), and typically, these can be combined simply 
to get a functioning path.  Thus, for metrics, the combining function 
is summation, and for bandwidth, the function is "min".

To accommodate attributes that are not carried as TE extensions today, 
we may have to add node attributes (say for dB loss across a mirror 
complex).  We may also have to linearize impairments, or define how 
they are composed across links and nodes; this is not always easy to do.

You say "fast allocating new wavelengths".  I would have thought that 
the issue is not how fast the allocation is, but whether the 
allocation and ultimately, the optical LSP setup, is successful taking 
into account optics and physics.

Kireeti.
-------