RE: New draft on wavelength switched optical networks

[Young Lee <ylee@huawei.com>]

Hi Snigdho,

Please see in-line for my comments.  Thanks.

Young

-----Original Message-----
From: owner-ccamp@ops.ietf.org [mailto:owner-ccamp@ops.ietf.org] On Behalf
Of Greg Bernstein
Sent: Saturday, July 14, 2007 10:46 AM
To: Bardalai, Snigdho
Cc: ccamp; Young Lee; pce@ietf.org
Subject: Re: New draft on wavelength switched optical networks

Hi Snigdho, good points and questions.  See comments below.

Regards

Greg B.

Bardalai, Snigdho wrote:
> Hi Greg,
>
> I believe your ID has presented some of the key points regarding
wavelength routing.
>
> I think we are still missing a few other issues that may have to be
considered.
> 1. Constraints related to the configuration of the ROADM switching
elements. For
>    example, transponders could be pre-wired to a specific port on the
ROADM, and hence
>    restricting the wavelengths that could be routed to that transponder.
>   
--> Yes. We touched on this only a bit but this is very important. There 
is a new draft (July 9, 2007) by Wataru. Imajuku, "Routing Extensions to 
Support Network Elements with Switching Constraint", 
draft-imajuku-ccamp-rtg-switching-constraint-02.txt. Which also hits 
some of these issues.  But this is an area that needs further 
requirements analysis.
It seems like we have at least:
(a) Internal switching topology constraints.  Such as you can't get to 
that port from this port. Illustrated in Wataru's draft.
(b) "Colored" interface related constraints where specific lambdas 
ingressing on a port will egress on a fixed port (not configurable). 
Like what you mention above.
(c) Wavelength converter based constraints such as we mention in our draft.
(d) ... Others? Or a better taxonomy than the above?

[Young] Agree with Greg. Wataru draft addressed the need to differentiate
interface types: (i) colored vs. (ii) colorless. 


> 2. When considering wavelength routing it may be important to consider
>    if regeneration of the signal is required. 
--> This kind of work was started by John Strand and Angela Chiu in 
RFC4054 on optical impairments related to routing.  Now since the 
publication the ITU-T has made a lot of progress in defining and 
characterizing various optical impairments so the time maybe about right 
to related some of this data plane work to the control plane. We 
originally were looking at this then saw some other gaps that needed 
filling.

[Young] The approach we have taken in regards to impairment issues in
wavelength optical switched network was to put on hold for now until we have
received enough interest in the current work. We (Greg and I) judged that
basic signaling and routing of the wavelengths should get kicked off before
we address optical impairment issues. 

But as you indicated, optical impairment issue is one of the key routing
constraints especially in the transparent optical network. We have not
forgotten this issue; but at this juncture, we'd like to pursue the issues
around the basic RWA issue first. Once this work is accepted in the
community, then we should pursue impairment issue. 


> Also, it may be equally important to
>    be able to specify, if and where reqeneration would be required during
signaling
>    (assuming an external entity such as a PCE can determine where the
regeneration can
>    be done).
>   
--> Yes.  We need regeneration capability information with our topology 
information which affects routing. Don't know that we'd need extensions 
to signaling, since once you've specified in the ERO to go through a 
regenerator element then you're done. At least for the fixed 
regenerators and those implicit in OEO switches.
>    

[Young]  One thing we should be careful, though, is routing scalability.
Previous attempts in this work have failed due to routing scalability issues
associated with the sheer amount of data that need to be advertised. But now
due to advancement of PCE, some of the information can be made available in
PCE (not necessarily via IGP) and PCE would handle path computation
constraints associated with regeneration and other optical impairment data. 

But before we jump into architectural alternatives, we should reach to an
agreement on the scope of essential data required to enable RWA. 

> It would be of much interest to me to learn what is your (and others)
opinion on these
> issues.
>
> Regards,
> Snigdho
>
> -----Original Message-----
> From: owner-ccamp@ops.ietf.org [mailto:owner-ccamp@ops.ietf.org]On
> Behalf Of Greg Bernstein
> Sent: Wednesday, June 27, 2007 12:39 PM
> To: ccamp; pce@ietf.org
> Cc: Young Lee
> Subject: New draft on wavelength switched optical networks
>
>
> Hi CCAMPer's and PCEr's, we have just published a new draft on the 
> "Applicability of GMPLS and PCE to Wavelength Switched Optical 
> Networks"  
>
http://www.ietf.org/internet-drafts/draft-bernstein-ccamp-wavelength-switche
d-00.txt 
> .
>
> This draft looks at optical networks that include tunable lasers and 
> ROADM (reconfigurable optical add/drop multiplexers) with no or limited 
> wavelength conversion capability (these components are defined in the 
> draft). 
> These limitations lead to the RWA (routing and wavelength assignment) 
> problem which is a bit more demanding in terms of input information and 
> computation than other constrained path computation problems.  In the 
> draft we look at the implications for GMPLS signaling, GMPLS routing, 
> and PCE protocols and suggest some potential extensions to better 
> accommodate this application.
>
> We'd appreciate feedback/collaboration on (a) overall interest in this 
> application, (b) requirements discussions, and (c) solution/extension 
> discussions.
>
> Cheers
>
> Greg B.
>
>   

-- 
===================================================
Dr Greg Bernstein, Grotto Networking (510) 573-2237