On Labels for Wavelength Switched Optical Networks...

[Greg Bernstein <gregb@grotto-networking.com>]

Hi folks, at the Vancouver meeting the lambda label format of Otani, et. al. draft-otani-ccamp-gmpls-lambda-labels-01.txt [Otani] was presented for a second time and a few new issues seemed to arise. I'm writing this note to see if we can resolve these on the list so this work can move forward, since the label format is valuable, in general, to the control of wavelength switched optical networks (WSON).

First, a general 32 bit lambda label has been defined in RFC3471.  This previous label does not directly relate to either the wavelength or frequency of the light used in a lambda LSP in a standardized way (folks can use the 32 bits as they see fit).
To come up with a completely general "lambda label" one could/should define both a (i) wavelength label specified in meters and (ii) a frequency label specified in Hertz (Hz).  These could be specified either with a 32 bit IEEE floating point number or via a suitable 32 bit integer by suitably adjusting the base units.  We could represent the frequency via a 32 bit integer in MHz, then a 193.1THz light source could be characterized by the integer 193,100,000. Similarly we could represent the wavelength label via a 32 bit integer in pico meters (10-12 meters), then a 1550nm wavelength could be characterized by the integer 1,550,000.

 Now any of the previous formats could be used with the 32 bit lambda label already defined.  The problem here is to pick a format for interoperability and compatibility with potentially common wavelength switched control operations.
Issues with the previously mentioned formats:

(a)    While floating point numbers provide great flexibility, as a label they have issues when it comes to comparison operations. 

(b)   An integer format with a suitably scaled exponent is relatively simple and just leaves the choice of “exponent” to be decided.

(c)    Neither format contains any “context” information about the WDM system in general.

The format proposed in [Otani]. avoids the above three issues and enhances common control plane operations as follows:

(a)    The format is integer based, hence avoids issues with floating point comparisons.

(b)   The format is based on the widely recognized ITU-T standard grids (ITU-T G.694.1 and .2) and fosters interoperability more than potentially any other choice. 

(c)    The ITU-T grids come in various widths and hence have an inherent growth path.

(d)   The ITU-T grids come in both frequency (G.694.1) and wavelength (G.694.2) flavored versions an both are incorporated in the [Otani] label format.

(e)    The format includes information on the grid spacing which is important WDM context information useful in many label selection processes.  For example a tunable laser associated with a 50GHz spacing WDM system could specify acceptable label range via the inclusive range label set mechanism.  Note that only those frequencies (labels) that fall on the grid are supported and not intermediate frequencies.

 
At the CCAMP WG meeting in Vancouver it was pointed out (by Lou Berger) that since a lambda label already exists and that existing implementations may make use of it that the proposed label of [Otani] would be better off referred to as a “G.694 label”. With such a change I think that this label format (and accompanying draft) should move forward as a working group document.

Comments, suggestions, issues?

Regards

Greg B.

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