Network Working Group                                G. Bernstein (ed.)
Internet Draft                                        Grotto Networking
Intended status: Informational                             Y. Lee (ed.)
Expires: August 2008                                             Huawei
                                                         Wataru Imajuku
                                                                    NTT




                                                      February 19, 2008

    Framework for GMPLS and PCE Control of Wavelength Switched Optical
                                 Networks
             draft-bernstein-ccamp-wavelength-switched-03.txt


Status of this Memo

   By submitting this Internet-Draft, each author represents that
   any applicable patent or other IPR claims of which he or she is
   aware have been or will be disclosed, and any of which he or she
   becomes aware will be disclosed, in accordance with Section 6 of
   BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups.  Note that
   other groups may also distribute working documents as Internet-
   Drafts.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   The list of current Internet-Drafts can be accessed at
   http://www.ietf.org/ietf/1id-abstracts.txt

   The list of Internet-Draft Shadow Directories can be accessed at
   http://www.ietf.org/shadow.html

   This Internet-Draft will expire on August 19, 2008.

Copyright Notice

   Copyright (C) The IETF Trust (2008).

Abstract



Bernstein and Lee      Expires August 19, 2008                 [Page 1]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   This memo provides a framework for applying Generalized Multi-
   Protocol Label Switching (GMPLS) and the Path Computation Element
   (PCE) architecture to the control of wavelength switched optical
   networks (WSON).  In particular we provide control plane models for
   key wavelength switched optical network subsystems and processes. The
   subsystems include wavelength division multiplexed links, tunable
   laser transmitters, reconfigurable optical add/drop multiplexers
   (ROADM) and wavelength converters.

   Lightpath provisioning, in general, requires the routing and
   wavelength assignment (RWA) process. This process is reviewed and the
   information requirements, both static and dynamic for this process
   are presented, along with alternative implementation scenarios that
   could be realized via GMPLS/PCE and/or extended GMPLS/PCE protocols.
   This memo does NOT address optical impairments in any depth and
   focuses on topological elements and path selection constraints that
   are common across different WSON environments.  It is expected that a
   variety of different techniques will be applied to optical
   impairments depending on the type of WSON, such as access, metro or
   long haul.



Conventions used in this document

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in RFC-2119 [RFC2119].

Table of Contents


   1. Introduction...................................................3
   2. Terminology....................................................4
   3. Wavelength Switched Optical Networks...........................5
      3.1. WDM and CWDM Links........................................5
      3.2. Optical Transmitters......................................7
         3.2.1. Lasers...............................................7
         3.2.2. Spectral Characteristics & Modulation Type...........8
         3.2.3. Signal Rates and Error Correction....................9
      3.3. ROADMs, OXCs, Splitters, Combiners and FOADMs............10
         3.3.1. Reconfigurable Add/Drop Multiplexers and OXCs.......10
         3.3.2. Splitters...........................................11
         3.3.3. Combiners...........................................12
         3.3.4. Fixed Optical Add/Drop Multiplexers.................12
      3.4. Wavelength Converters....................................13
   4. Routing and Wavelength Assignment.............................15
      4.1. Lightpath Temporal Characteristics.......................16


   Bernstein and Lee Expires August 19, 2008 [Page 2]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


      4.2. RWA Algorithmic Approaches: Combined vs. Two-step........17
      4.3. RWA Computation Architectures............................17
      4.4. Conveying information needed by RWA......................18
   5. GMPLS & PCE Implications......................................19
      5.1. Implications for GMPLS signaling.........................19
         5.1.1. Identifying Wavelengths and Signals.................19
         5.1.2. Combined RWA/Separate Routing WA support............20
         5.1.3. Distributed Wavelength Assignment: Unidirectional, No
         Converters.................................................20
         5.1.4. Distributed Wavelength Assignment: Unidirectional,
         Limited Converters.........................................21
         5.1.5. Distributed Wavelength Assignment: Bidirectional, No
         Converters.................................................21
      5.2. Implications for GMPLS Routing...........................21
         5.2.1. Need for Wavelength-Specific Maximum Bandwidth
         Information................................................22
         5.2.2. Need for Wavelength-Specific Availability Information22
         5.2.3. Relationship to Link Bundling and Layering..........23
         5.2.4. WSON Routing Information Summary....................23
      5.3. Optical Path Computation and Implications for PCE........25
         5.3.1. Lightpath Constraints and Characteristics...........25
         5.3.2. Computation Architecture Implications...............26
         5.3.3. Discovery of RWA Capable PCEs.......................26
      5.4. Scaling Implications.....................................26
         5.4.1. Routing.............................................27
         5.4.2. Signaling...........................................27
         5.4.3. Path computation....................................27
   6. Security Considerations.......................................27
   7. IANA Considerations...........................................27
   8. Acknowledgments...............................................27
   9. References....................................................29
      9.1. Normative References.....................................29
      9.2. Informative References...................................30
   10. Contributors.................................................32
   Author's Addresses...............................................32
   Intellectual Property Statement..................................33
   Disclaimer of Validity...........................................34

1. Introduction

   From its beginning Generalized Multi-Protocol Label Switching (GMPLS)
   was intended to control wavelength switched optical networks (WSON)
   with the GMPLS architecture document [RFC3945] explicitly mentioning
   both wavelength and waveband switching and equating wavelengths
   (lambdas) with GMPLS labels. In addition a discussion of optical
   impairments and other constraints on optical routing can be found in
   [RFC4054]. However, optical technologies have advanced in ways that
   make them significantly different from other circuit switched


   Bernstein and Lee Expires August 19, 2008 [Page 3]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   technologies such as Time Division Multiplexing (TDM). Service
   providers have already deployed many of these new optical
   technologies such as ROADMs and tunable lasers and desire the same
   automation and restoration capabilities that GMPLS has provided to
   TDM and packet switched networks. Another important application of an
   automated control plane such as GMPLS is the possibility to improve,
   via recovery schemes, the availability of the network.  One of the
   key points of GMPLS based recovery schemes is the capability to
   survive multiple failures while legacy protection mechanism such as
   1+1 path protection can survive to a single failure.  Moreover this
   improved availability can be obtained using less network resources.

   This document will focus on the unique properties of links, switches
   and path selection constraints that occur in WSONs.  Different WSONs
   such as access, metro and long haul may apply different techniques
   for dealing with optical impairments hence this document will NOT
   address optical impairments in any depth, but instead focus on
   properties that are common across a variety of WSONs.

   This memo provides a framework for applying GMPLS and the Path
   Computation Element (PCE) architecture to the control of WSONs.  In
   particular we provide control plane models for key wavelength
   switched optical network subsystems and processes. The subsystems
   include wavelength division multiplexed links, tunable laser
   transmitters, reconfigurable optical add/drop multiplexers (ROADM)
   and wavelength converters.

   Lightpath provisioning, in general, requires the routing and
   wavelength assignment (RWA) process. This process is reviewed and the
   information requirements, both static and dynamic for this process
   are presented, along with alternative implementation architectures
   that could be realized via various combinations of extended GMPLS and
   PCE protocols.

2. Terminology

   CWDM: Coarse Wavelength Division Multiplexing.

   DWDM: Dense Wavelength Division Multiplexing.

   FOADM: Fixed Optical Add/Drop Multiplexer.

   OXC: Optical cross connect. A symmetric optical switching element in
   which a signal on any ingress port can reach any egress port.

   ROADM: Reconfigurable Optical Add/Drop Multiplexer. An asymmetric
   wavelength selective switching element featuring ingress and egress
   line side ports as well as add/drop side ports.


   Bernstein and Lee Expires August 19, 2008 [Page 4]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   RWA: Routing and Wavelength Assignment.

   Wavelength Conversion/Converters: The process of converting an
   information bearing optical signal centered at a given wavelength to
   one with "equivalent" content centered at a different wavelength.
   Wavelength conversion can be implemented via an optical-electronic-
   optical (OEO) process or via a strictly optical process.

   WDM: Wavelength Division Multiplexing.

   Wavelength Switched Optical Networks (WSON): WDM based optical
   networks in which switching is performed selectively based on the
   center wavelength of an optical signal.

3. Wavelength Switched Optical Networks

   WSONs come in a variety of shapes and sizes from continent spanning
   long haul networks, to metropolitan networks, to residential access
   networks. In all these cases we are concerned with those properties
   that constrain the choice of wavelengths that can be used, i.e.,
   restrict the wavelength label set, impact the path selection process,
   and limit the topological connectivity. To do so we will examine and
   model some major subsystems of a WSON: WDM links, Optical
   Transmitters, ROADMs, and Wavelength Converters.

3.1. WDM and CWDM Links

   WDM and CWDM links run over optical fibers, and optical fibers come
   in a wide range of types that tend to be optimized for various
   applications from access networks, metro, long haul, and submarine
   links to name a few. ITU-T and IEC standards exist for various types
   of fibers. For the purposes here we are concerned only with single
   mode fibers (SMF). The following SMF fiber types are typically
   encountered in optical networks:

      ITU-T Standard |  Common Name
      ------------------------------------------------------------
      G.652 [G.652]  |  Standard SMF                              |
      G.653 [G.653]  |  Dispersion shifted SMF                    |
      G.654 [G.654]  |  Cut-off shifted SMF                       |
      G.655 [G.655]  |  Non-zero dispersion shifted SMF           |
      G.656 [G.656]  |  Wideband non-zero dispersion shifted SMF  |
      ------------------------------------------------------------
   These fiber types are differentiated by their optical impairment
   characteristics such as attenuation, chromatic dispersion,
   polarization mode dispersion, four wave mixing, etc. Since these
   effects can be dependent upon wavelength, channel spacing and input



   Bernstein and Lee Expires August 19, 2008 [Page 5]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   power level, the net effect for our modeling purposes here is to
   restrict the range of wavelengths that can be used.

   Typically WDM links operate in one or more of the approximately
   defined optical bands [G.Sup39]:

      Band     Range (nm)     Common Name    Raw Bandwidth (THz)
      O-band   1260-1360      Original       17.5
      E-band   1360-1460      Extended       15.1
      S-band   1460-1530      Short          9.4
      C-band   1530-1565      Conventional   4.4
      L-band   1565-1625      Long           7.1
      U-band   1625-1675      Ultra-long     5.5



   Not all of a band may be usable, for example in many fibers that
   support E-band there is significant attenuation due to a water
   absorption peak at 1383nm. Hence we can have a discontinuous
   acceptable wavelength range for a particular link. Also some systems
   will utilize more than one band. This is particularly true for coarse
   WDM (CWDM) systems.

   [Editor's note: the previous text is primarily tutorial in nature and
   maybe deleted or moved to an appendix in a future draft]

   Current technology breaks up the bandwidth capacity of fibers into
   distinct channels based on either wavelength or frequency. There are
   two standards covering wavelengths and channel spacing. ITU-T
   recommendation [G.694.1] describes a DWDM grid defined in terms of
   frequency grids of 12.5GHz, 25GHz, 50GHz, 100GHz, and other multiples
   of 100GHz around a 193.1THz center frequency. At the narrowest
   channel spacing this provides less than 4800 channels across the O
   through U bands. ITU-T recommendation [G.694.2] describes a CWDM grid
   define in terms of wavelength increments of 20nm running from 1271nm
   to 1611nm for 18 or so channels. The number of channels is
   significantly smaller than the 32 bit GMPLS label space allocated to
   lambda switching.  A fixed mapping between the GMPLS label space and
   these ITU-T WDM grids as proposed in [Otani] would not only allow a
   common vocabulary to be used in signaling lightpaths but also in
   describing WDM links, ROADM ports, and wavelength converters for the
   purposes path selection.

   With a tremendous existing base of fiber many WDM links are designed
   to take advantage of particular fiber characteristics or to try to
   avoid undesirable properties.  For example dispersion shifted SMF
   [G.653] was originally designed for good long distance performance in
   single channel systems, however putting WDM over this type of fiber


   Bernstein and Lee Expires August 19, 2008 [Page 6]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   requires much system engineering and a fairly limited range of
   wavelengths. Hence for our basic, impairment unaware, modeling of a
   WDM link we will need the following information:

   o  Wavelength range(s): Given a mapping between labels and the ITU-T
      grids each range could be expressed in terms of a doublet
      (lambda1, lambda2) or (freq1, freq1) where the lambdas or
      frequencies can be represented by 32 bit integers.

   o  Channel spacing: currently there are about five channel spacings
      used in DWDM systems 12.5GHz to 200GHz and one defined CWDM
      spacing.

   For a particular link this information is relatively static, i.e.,
   changes to these properties generally require hardware upgrades. Such
   information could be used locally during wavelength assignment via
   signaling, similar to label restrictions in MPLS or used by a PCE in
   solving the combined routing and wavelength assignment problem.

3.2. Optical Transmitters

   3.2.1. Lasers

   WDM optical systems make use of laser transmitters utilizing
   different wavelengths (frequencies). Some laser transmitters were and
   are manufactured for a specific wavelength of operation, that is, the
   manufactured frequency cannot be changed. First introduced to reduce
   inventory costs, tunable optical laser transmitters are becoming
   widely deployed in some systems [Coldren04], [Buus06]. This allows
   flexibility in the wavelength used for optical transmission and aids
   in the control of path selection.

   Fundamental modeling parameters from the control plane perspective
   optical transmitters are:

   o  Tunable: Is this transmitter tunable or fixed.

   o  Tuning range: This is the frequency or wavelength range over which
      the laser can be tuned. With the fixed mapping of labels to
      lambda's of [Otani] this can be expressed as a doublet (lambda1,
      lambda2) or (freq1, freq2) where lambda1 and lambda2 or freq1 and
      freq2 are the labels representing the lower and upper bounds in
      wavelength or frequency.







   Bernstein and Lee Expires August 19, 2008 [Page 7]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   o  Tuning time: Tuning times highly depend on the technology used.
      Thermal drift based tuning may take seconds to stabilize, whilst
      electronic tuning might provide sub-ms tuning times. Depending on
      the application this might be critical. For example, thermal drift
      might not be applicable for fast protection applications.

   o  Spectral Characteristics and stability: The spectral shape of the
      laser's emissions and its frequency stability put limits on
      various properties of the overall WDM system. One relatively easy
      to characterize constraint is the finest channel spacing on which
      the transmitter can be used.

   Note that ITU-T recommendations specify many other aspects of a
   laser's such as spectral characteristics and stability. Many of these
   parameters a key in designing WDM subsystems consisting of
   transmitters, WDM links and receivers however they do not furnish
   additional information that will influence label switched path (LSP)
   provisioning in a properly designed system.

   Also note that lasers transmitters as a component can degrade and
   fail over time. This presents the possibility of the failure of a LSP
   (lightpath) without either a node or link failure. Hence, additional
   mechanisms may be necessary to detect and differentiate this failure
   from the others, e.g., one doesn't not want to initiate mesh
   restoration if the source transmitter has failed, since the laser
   transmitter will still be failed on the alternate optical path.

   3.2.2. Spectral Characteristics & Modulation Type

   Contrary to some marketing claims optical systems are not truly
   "transparent" to the content of the signals that they carry. Each
   lightpath will have spectral characteristics based on its content,
   and the spacing of wavelengths in a WDM link will ultimately put
   constraints on that spectrum.

   For analog signals such as used in closed access television (CATV) or
   "radio over fiber" links spectral characteristics are given in terms
   of various bandwidth measures. However digital signals consist of our
   main focus here and in the ITU-T G series optical specifications. In
   this case the spectral characteristics can be more accurately
   inferred from the modulation format and the bit rate.

   Although Non-Return to Zero (NRZ) is currently the dominant form of
   optical modulation, new modulation formats are being researched
   [Winzer06] and deployed. With a choice in modulation formats we no
   longer have a one to one relationship between digital bandwidth in
   bytes or bits per second and the amount of optical spectrum (optical
   bandwidth) consumed. To simplify the specification of optical signals


   Bernstein and Lee Expires August 19, 2008 [Page 8]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   the ITU-T, in recommendation G.959.1, combined a rate bound and
   modulation format designator [G.959.1]. For example, two of the
   signal classes defined in [G.959.1] are:

   Optical tributary signal class NRZ 1.25G:

     "Applies to continuous digital signals with non-return to zero line
     coding, from nominally 622 Mbit/s to nominally 1.25 Gbit/s. Optical
     tributary signal class NRZ 1.25G includes a signal with STM-4 bit
     rate according to ITU-T Rec. G.707/Y.1322." Note that Gigabit
     Ethernet falls into this signaling class as well.

   Optical tributary signal class RZ 40G:

     "Applies to continuous digital signals with return to zero line
     coding, from nominally 9.9 Gbit/s to nominally 43.02 Gbit/s.
     Optical tributary signal class RZ 40G includes a signal with STM-
     256 bit rate according to ITU-T Rec. G.707/Y.1322 and OTU3 bit rate
     according to ITU-T Rec. G.709/Y.1331."

   From a modeling perspective we have:

   o  Analog signals: bandwidth parameters, e.g., 3dB parameters and
      similar.

   o  Digital signals: there are predefined modulation bit rate classes
      that we can encode.

   This information can be important in constraining route selection,
   for example some signals may not be compatible with some links or
   wavelength converters. In addition it lets the endpoints understand
   if it can process the signal.

   3.2.3. Signal Rates and Error Correction

   Although, the spectral characteristics of a signal determine its
   basic compatibility with a WDM system, more information is generally
   needed for various processing activities such as regeneration and
   reception. Many digital signals such as Ethernet, G.709, and SDH have
   well defined encoding which includes forward error correction (FEC).
   However many subsystem vendors offer additional FEC options for a
   given signal type. The use of different FECs can lead to different
   overall signal rates. If the FEC and rate used is not compatible
   between the sender and receiver the signal can not be correctly
   processed. Note that the rates of "standard" signals may be extended
   to accommodate different payloads.  For example there are
   transmitters capable of directly mapping 10GE LAN-PHY traffic into
   G.709 ODU2 frame with slightly higher clock rate [G.Sup43].


   Bernstein and Lee Expires August 19, 2008 [Page 9]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


3.3. ROADMs, OXCs, Splitters, Combiners and FOADMs

   3.3.1. Reconfigurable Add/Drop Multiplexers and OXCs

   Reconfigurable add/drop optical multiplexers (ROADM) have matured and
   are available in different forms and technologies [Basch06]. This is
   a key technology that allows wavelength based optical switching. A
   classic degree-2 ROADM is shown in Figure 1.


        Line side ingress    +---------------------+  Line side egress
                         --->|                     |--->
                             |                     |
                             |        ROADM        |
                             |                     |
                             |                     |
                             +---------------------+
                                 | | | |  o o o o
                                 | | | |  | | | |
                                 O O O O  | | | |
         Tributary Side:   Drop (egress)  Add (ingress)

                          Figure 1 Degree-2 ROADM

   The key feature across all ROADM types is their highly asymmetric
   switching capability. In the ROADM of Figure 1, the "add" ingress
   ports can only egress on the line side egress port and not on any of
   the "drop" egress ports. The degree of a ROADM or switch is given by
   the number of line side ports (ingress and egress) and does not
   include the number of "add" or "drop" ports. Sometimes the "add"
   "drop" ports are also called tributary ports. As the degree of the
   ROADM increases beyond two it can have properties of both a switch
   (OXC)and a multiplexer and hence we must know the potential
   connectivity offered by such a network element to effectively utilize
   it. A straight forward way to do this is via a "potential
   connectivity" matrix A where Amn = 0 or 1, depending upon whether a
   wavelength on ingress port m can be connected to egress port n
   [Imajuku]. For the ROADM of Figure 1 the potential connectivity
   matrix can be expressed as











   Bernstein and Lee Expires August 19, 2008 [Page 10]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


               Ingress  Egress Port
               Port     #1 #2 #3 #4 #5
                        --------------
               #1:      1  1  1  1  1
               #2       1  0  0  0  0
         A =   #3       1  0  0  0  0
               #4       1  0  0  0  0
               #5       1  0  0  0  0

   Where ingress ports 2-5 are add ports, egress ports 2-5 are drop
   ports and ingress port #1 and egress port #1 are the line side (WDM)
   ports.

   For ROADMs this matrix will be very sparse, and for OXCs the
   complement of the matrix will be very sparse hence even relatively
   high degree ROADMs/OXCs can be compactly characterized.

   Additional constraints may also apply to the various ports in a
   ROADM/OXC. In the literature of optical switches and ROADMs the
   following restrictions/terms are used:

   Colored port: An ingress or more typically an egress (drop) port
   restricted to a single channel of fixed wavelength.

   Colorless port: An ingress or more typically an egress (drop) port
   restricted to a single channel of arbitrary wavelength.

   In general a port on a ROADM could have any of the following
   wavelength restrictions:

   o  Multiple wavelengths, full range port

   o  Single wavelength, full range port

   o  Single wavelength, fixed lambda port

   o  Multiple wavelengths, reduced range port (like wave band
      switching)

   To model these restrictions we need two pieces of information for
   each port: (a) number of wavelengths, (b) wavelength range and
   spacing.  Note that this information is relatively static.

   3.3.2. Splitters

   An optical splitter consists of a single ingress port and two or more
   egress ports. The ingress optical signaled is essentially copied
   (with loss) to all egress ports.


   Bernstein and Lee Expires August 19, 2008 [Page 11]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   Using the modeling notions of section 3.3.1. the ingress and egress
   ports of a splitter would have the same wavelength restrictions. In
   addition we can describe a splitter by a connectivity matrix Amn as
   follows:

               Ingress  Egress Port
               Port     #1 #2 #3 ...   #N
                        -----------------
         A =   #1       1  1  1  ...   1

   The difference from a simple ROADM is that this is not a "potential"
   connectivity matrix but the fixed connectivity of the device. Hence
   to differentiate between a simple ROADM and a splitter we'd need an
   indicator as to whether the device is reconfigurable.

   3.3.3. Combiners

   A optical combiner is somewhat the dual of a splitter in that it has
   a single multi-wavelength egress port and multiple ingress ports.
   The contents of all the ingress ports are copied and combined to the
   single egress port.  The various ports may have different wavelength
   restrictions. It is generally the responsibility of those using the
   combiner to assure that wavelength collision does not occur on the
   egress port. The connectivity matrix Amn for a combiner would look
   like:

               Ingress  Egress Port
               Port     #1
                        ---
               #1:      1
               #2       1
         A =   #3       1
               ...      1
               #N       1

   Once again this connectivity is fixed.

   3.3.4. Fixed Optical Add/Drop Multiplexers

   A fixed optical add/drop multiplexer can alter the course of an
   ingress wavelength in a preset way. In particular a particular
   wavelength (or waveband) from a line side ingress port would be
   dropped to a particular "tributary" egress port. Depending on the
   device's fixed configuration that same wavelength may or may not be
   "continued" to the line side egress port ("drop and continue"
   operation).  Further there may exist tributary ingress ports ("add"
   ports) whose signals are combined with each other and "continued"
   line side signals.


   Bernstein and Lee Expires August 19, 2008 [Page 12]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   In general to represent the routing properties of an FOADM we need
   the connectivity matrix Amn as previously discussed and we need the
   precise wavelength restrictions for all ingress and egress ports.
   From the wavelength restrictions on the tributary egress ports (drop
   ports) we can see what wavelengths have been dropped. From the
   wavelength restrictions on the tributary ingress (add) ports we can
   see which wavelengths have been added to the line side egress port.
   Finally from the added wavelength information and the line side
   egress wavelength restrictions we can infer which wavelengths have
   been continued.

   To summarize, the modeling methodology introduced in section 3.3.1.
   consisting of a connectivity matrix and port wavelength restrictions
   can be used to describe a large set of fixed optical devices such as
   combiners, splitters and FOADMs with the inclusion of an indicator as
   to whether the device is Fixed or Reconfigurable and the appropriate
   interpretations detailed previously.

3.4. Wavelength Converters

   Wavelength converters take an ingress optical signal at one
   wavelength and emit an equivalent content optical signal at another
   wavelength on egress. There are currently two approaches to building
   wavelength converters. One approach is based on optical to electrical
   to optical (OEO) conversion with tunable lasers on egress. This
   approach can be dependent upon the signal rate and format, i.e., this
   is basically an electrical regenerator combined with a tunable laser.
   The other approach performs the wavelength conversion, optically via
   non-linear optical effects, similar in spirit to the familiar
   frequency mixing used in radio frequency systems, but significantly
   harder to implement.  Such processes/effects may place limits on the
   range of achievable conversion. These may depend on the wavelength of
   the input signal and the properties of the converter as opposed to
   the only the properties of the converter in the OEO case. Different
   WSON system designs may choose to utilize this component to varying
   degrees or not at all.

   Current or envisioned contexts for wavelength converters are:

  1. Wavelength conversion associated with OEO switches and tunable
     laser transmitters. In this case there are plenty of converters to
     go around since we can think of each tunable output laser
     transmitter on an OEO switch as a potential wavelength converter.

  2. Wavelength conversion associated with ROADMs/OXCs. In this case we
     may have a limited amount of conversion available. Conversion could
     be either all optical or via an OEO method.



   Bernstein and Lee Expires August 19, 2008 [Page 13]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


  3. Wavelength conversion associated with fixed devices such as FOADMs.
     In this case we may have a limited amount of conversion. Also in
     this case the conversion may be used as part of light path routing.

   Based on the above contexts a tentative modeling approach for
   wavelength converters could be as follows:

   1. Wavelength converters can always be modeled as associated with
      network elements. This includes fixed wavelength routing elements.

   2. A network element may have full wavelength conversion capability,
      i.e., any ingress port and wavelength, or a limited number of
      wavelengths and ports. On a box with a limited number of
      converters there also may exist restrictions on which ports can
      reach the converters. Hence regardless of where the converters
      actually are we can associate them with ingress ports.

   3. Wavelength converters have range restrictions that are either
      independent or dependent upon the ingress wavelength. [TBD: for
      those that depend on ingress wavelength can we have a standard
      formula? Also note that this type of converter introduces
      additional optical impairments.]

   4. Wavelength converters that are O-E-O based will have a restriction
      based on the modulation format and transmission speed.

   Note that since O-E-O wavelength converters also serve as
   regenerators we can include regenerators in our model of wavelength
   converters. O-E-O Regenerators come in three general types known as
   1R, 2R, and 3R regenerators. 1R regenerators re-amplify the signal to
   combat attenuation, 2R regenerators reshape as well as amplify the
   signal, 3R regenerators amplify, reshape and retime the signal. As we
   go from 1R to 3R regenerators the signal is ''cleaned up'' better but
   at the same time the regeneration process becomes more dependent on
   the signal characteristics such as format and rate.

   In WSONs where wavelength converters are sparse we may actually see a
   light path appear to loop or ''backtrack'' upon itself in order to
   reach a wavelength converter prior to continuing on to its
   destination. The lambda used on the "detour" out to the wavelength
   converter would be different from that coming back from the "detour"
   to the wavelength converter.

   A model for an O-E-O wavelength converter would consist of:

   o  Input lambda or frequency range

   o  Output lambda or frequency range


   Bernstein and Lee Expires August 19, 2008 [Page 14]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   o  Equivalent regeneration level (1R, 2R, 3R)

   o  Signal restrictions if a 2R or 3R regeneration: formats and rates

  [FFS: Model for an all optical wavelength converter]

4. Routing and Wavelength Assignment

   In wavelength switched optical networks consisting of tunable lasers
   and wavelength selective switches with wavelength converters on every
   interface, path selection is similar to the MPLS and TDM circuit
   switched cases in that the labels, in this case wavelengths
   (lambdas), have only local significance. That is, a wavelength-
   convertible network with full wavelength-conversion capability at
   each node is equivalent to a circuit-switched TDM network with full
   time slot interchange capability; thus, the routing problem needs to
   be addressed only at the level of the traffic engineered (TE) link
   choice, and wavelength assignment can be resolved locally by the
   switches on a hop-by-hop basis.

   However, in the limiting case of an optical network with no
   wavelength converters, a light path (optical channel - OCh -) needs a
   route from source to destination and must pick a single wavelength
   that can be used along that path without "colliding" with the
   wavelength used by any other light path that may share an optical
   fiber. This is sometimes referred to as a "wavelength continuity
   constraint". To ease up on this constraint while keeping network
   costs in check a limited number of wavelength converters maybe
   introduce at key points in the network [Chu03].

   In the general case of limited or no wavelength converters this
   computation is known as the Routing and Wavelength Assignment (RWA)
   problem [HZang00]. The "hardness" of this problem is well documented.
   There, however, exist a number of reasonable approximate methods for
   its solution [HZang00].

   The inputs to the basic RWA problem are the requested light paths
   source and destination, the networks topology, the locations and
   capabilities of any wavelength converters, and the wavelengths
   available on each optical link. The output from an algorithm solving
   the RWA problem is an explicit route through ROADMs, a wavelength for
   the optical transmitter, and a set of locations (generally associated
   with ROADMs or switches) where wavelength conversion is to occur and
   the new wavelength to be used on each component link after that point
   in the route.

   It is to be noted that choice of specific RWA algorithm is out of the
   scope for this document. However there are a number of different


   Bernstein and Lee Expires August 19, 2008 [Page 15]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   approaches to dealing with the RWA algorithm that can affect the
   division of effort between signaling, routing and PCE.

4.1. Lightpath Temporal Characteristics

   Key inputs that can affect the choice of solution to the RWA process
   are those associated with the temporal characteristics of a light
   path connection. For our purposes here we look at the timeliness of
   connection establishment/teardown, and the duration of the
   connection.

   Connection Establishment/Teardown Timeliness can be thought of in
   approximately three time frames:

  1. Time Critical:  For example those lightpath establishments used for
     restoration of service or other high priority real time service
     requests.

  2. Soft time bounds: This is a more typical new connection request.
     While expected to be responsive, there should be more time to take
     into account network optimization.

  3. Scheduled or Advanced reservations. Here lightpath connections are
     requested significantly ahead of their intended "in service" time.
     There is the potential for significant network optimization if
     multiple lightpaths can be computed in parallel to achieve network
     optimization objectives.

  Lightpath connection duration has typically been thought of as
     approximately three time frames:

  1. Dynamic: those lightpaths with relatively short duration (holding
     times).

  2. Pseudo-static: lightpaths with moderately long durations.

  3. Static: lightpaths with long durations.

   Different types of RWA algorithms have been developed for dealing
   with dynamic versus pseudo-static conditions.  These can address
   service provider's needs for: (a) network optimization, (b)
   restoration, and (c) highly dynamic lightpath provisioning.

   Hence we can model timescale related lightpath requirements via the
   following notions:

   o  Batch or Sequential light path connection requests



   Bernstein and Lee Expires August 19, 2008 [Page 16]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   o  Timeliness of Connection establishment

   o  Duration of lightpath connection

4.2. RWA Algorithmic Approaches: Combined vs. Two-step

   When solving the RWA problem some algorithms that solve the problem
   in a two step procedure of path selection followed by wavelength
   assignment, and others that solve the problem in a combined fashion.
   These different types of algorithms can have different
   characteristics with respect to computational complexity and
   optimality that can influence how and where they may be used.

   Given a path, the following are a non-exhaustive subset of wavelength
   assignment (WA) approaches discussed in [HZang00]:

   1. Random: Looks at all available wavelengths for the light path then
      chooses from those available at random.

   2. First Fit: Wavelengths are ordered, first available (on all links)
      is chosen.

   3. Most Used: Out of the wavelengths available on the path attempts
      to select most use wavelength in network.

   4. Least Loaded: For multi-fiber networks. Chooses the wavelength j
      that maximizes minimum of the difference between the number of
      fibers on link l and the number of fibers on link l with
      wavelength j occupied.

   As can be seen from the above short list, wavelength assignment
   methods have differing information or processing requirements. When a
   two step RWA algorithm is used there is also the possibility of
   utilizing the signaling system to perform distributed wavelength
   assignment. The details would differ according to the WA algorithm
   chosen but at a minimum all would take an intersection of available
   wavelength sets as seen at each link along the path to produce the
   available wavelength set for the path.

4.3. RWA Computation Architectures

   The previous discussion leads to the following possible RWA
   computation architectures. As will be seen in the GMPLS/PCE
   evaluation section these architectures have different impacts on
   their needs with respect to protocol extensions.





   Bernstein and Lee Expires August 19, 2008 [Page 17]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   o  Combined RWA --- Both routing and wavelength assignment are
      performed at a single computational entity. This choice assumes
      that computational entity has sufficient WSON network link/nodal
      information and topology to be able to compute RWA. The amount of
      information can vary depending on the RWA approach/algorithm
      utilized. Good for network optimization and for smaller WSONs.

   o  Separate Routing and WA --- Separate entities perform routing and
      wavelength assignment. The path obtained from the routing
      computational entity must be furnished to the entity performing
      wavelength assignment. Good for offloading some of the
      computational burden and for reducing the number of entities that
      need exact network link wavelength utilization, i.e., there can be
      fewer nodes that perform WA than perform routing.

   o  Routing with Distributed WA --- Routing is performed at a
      computational entity while wavelength assignment is performed in a
      distributed fashion across nodes along the path. Good in that it
      does not require exact network link wavelength information at any
      node, but can have higher blocking probabilities compared to other
      methods.

4.4. Conveying information needed by RWA

   The previous sections have characterized WSONs and lightpath
   requests. In particular high level models of the information by the
   RWA process were presented. We can view this information as either
   static, changing with hardware changes (including possibly failures),
   or dynamic, can change with subsequent lightpath provisioning. The
   timeliness in which an entity involved in the RWA process is notified
   of such changes is fairly situational. For example, for network
   restoration purposes, learning of a hardware failure or of new
   hardware coming online to provide restoration capability can be
   critical.
   Currently there are various methods for communicating RWA relevant
   information, these include, but are not limited to:

   o  Existing control plane protocols such as GMPLS routing and
      signaling. Note that routing protocols can be used to convey both
      static and dynamic information. Static information currently
      conveyed includes items like router options and such.

   o  Management protocols such as NetConf, SNMPv3, CLI, CORBA, or
      others.

   o  Directory services and accompanying protocols. These are good for
      the dissemination of relatively static information. Not intended
      for dynamic information.


   Bernstein and Lee Expires August 19, 2008 [Page 18]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   o  Other techniques for dynamic information: messaging straight from
      NEs to PCE to avoid flooding. This would be useful if the number
      of PCEs is significantly less than number of WSON NEs. Or other
      ways to limit flooding to "interested" NEs.

   Mechanisms to improve scaling of dynamic information:

   o  Tailor message content to WSON. For example the use of wavelength
      ranges, or wavelength occupation bit maps.

   o  Utilize incremental updates if feasible.

5. GMPLS & PCE Implications

   The presence and amount of wavelength conversion available at a
   wavelength switching interface has an impact on the information that
   needs to be transferred by the control plane (GMPLS) and the PCE
   architecture. Current GMPLS and PCE standards can address the full
   wavelength conversion case so the following with only address the
   limited and no wavelength conversion cases.

5.1. Implications for GMPLS signaling

   Basic support for WSON signaling already exists in GMPLS with the
   lambda (value 9) LSP encoding type [RFC3471], or for G.709 compatible
   optical channels, the LSP encoding type (value = 13) "G.709 Optical
   Channel" from [RFC4328]. However a number of practical issues arise
   in the identification of wavelengths and signals, and distributed
   wavelength assignment processes which are discussed below.

   5.1.1. Identifying Wavelengths and Signals

   As previously stated a global fixed mapping between wavelengths and
   labels simplifies the characterization of WDM links and WSON devices.
   Furthermore such a mapping as described in [Otani] eases
   communication between PCE and WSON PCCs.

   An alternative to a global network map of labels to wavelengths would
   be to use LMP to assign the map for each link then convey that
   information to any path computation entities, e.g., label switch
   routers or stand alone PCEs. The local label map approach will
   require the label-set contents in the RSVP-TE Path message to be
   translated every time the map changes between an incoming link and
   the outgoing link.

   In the future, it maybe worthwhile to define traffic parameters for
   lambda LSPs that include a signal type field that includes modulation



   Bernstein and Lee Expires August 19, 2008 [Page 19]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   format/rate information. This is similar to what was done in
   reference [RFC4606] for SONET/SDH signal types.

   5.1.2. Combined RWA/Separate Routing WA support

   In either the combined RWA or separate routing WA cases, the node
   initiating the signaling will have a route from the source to
   destination along with the wavelengths (generalized labels) to be
   used along portions of the path. Current GMPLS signaling supports an
   explicit route object (ERO) and within an ERO an ERO Label subobject
   can be use to indicate the wavelength to be used at a particular
   node. In case the local label map approach is used the label sub-
   object entry in the ERO has to be translated appropriately.

   5.1.3. Distributed Wavelength Assignment: Unidirectional, No
      Converters

   GMPLS signaling for a uni-directional lightpath LSP allows for the
   use of a label set object in the RSVP-TE path message. The processing
   of the label set object to take the intersection of available lambdas
   along a path can be performed resulting in the set of available
   lambda being known to the destination that can then use a wavelength
   selection algorithm to choose a lambda. The algorithms that can be
   used would be limited to the information available at the destination
   node. The information requirements of the methods discussed in
   section 4.2. are as follows:

  1. Random: nothing more than the available wavelength set.

  2. First Fit: nothing more than the available wavelength set.

  3. Most Used: the available wavelength set and information on global
     wavelength use in the network.

  4. Least Loaded: the available wavelength set and information
     concerning the wavelength dependent loading for each link (this
     applies to multi-fiber links). This could be obtained via global
     information or via supplemental information passed via the
     signaling protocol.

   In case (3) above the global information needed by the wavelength
   assignment could be derived from suitably enhanced GMPLS routing.
   Note however this information need not be accurate enough for
   combined RWA computation. Currently, GMPLS signaling does not provide
   a way to indicate that a particular wavelength assignment algorithm
   should be used.




   Bernstein and Lee Expires August 19, 2008 [Page 20]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   5.1.4. Distributed Wavelength Assignment: Unidirectional, Limited
      Converters

   The previous outlined the case with no wavelength converters. In the
   case of wavelength converters nodes with wavelength converters would
   need to make the decision as to whether to perform conversion. One
   indicator for this would be that the set of available wavelengths
   which is obtained via the intersection of the incoming label set and
   the egress links available wavelengths is either null or deemed too
   small to permit successful completion.

   At this point the node would need to remember that it will apply
   wavelength conversion and will be responsible for assigning the
   wavelength on the previous lambda-contiguous segment when the RSVP-TE
   RESV message passes by. The node will pass on an enlarged label set
   reflecting only the limitations of the wavelength converter and the
   egress link. The record route option in RVSP-TE signaling can be used
   to show where wavelength conversion has taken place.

   5.1.5. Distributed Wavelength Assignment: Bidirectional, No
      Converters

   There are potential issues in the case of a bi-directional lightpath
   which requires the use of the same lambda in both directions. We can
   try to use the above procedure to determine the available
   bidirectional lambda set if we use the interpretation that the
   available label set is available in both directions. However, a
   problem, arises in that bidirectional LSPs setup, according to
   [RFC3471] section 4.1, is indicated by the presence of an upstream
   label in the path message.

   However, until the intersection of the available label sets is
   obtained, e.g., at the destination node and the wavelength assignment
   algorithm has been run the upstream label information will not be
   available. Hence currently distributed wavelength assignment with
   bidirectional lightpaths is not supported.


5.2. Implications for GMPLS Routing

   GMPLS routing [RFC4202] currently defines an interface capability
   descriptor for "lambda switch capable" (LSC) which we can use to
   describe the interfaces on a ROADM or other type of wavelength
   selective switch. In addition to the topology information typically
   conveyed via an IGP, we would need to convey the following subsystem
   properties to minimally characterize a WSON:

  1. WDM Link properties (allowed wavelengths).


   Bernstein and Lee Expires August 19, 2008 [Page 21]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


  2. Laser Transmitters (wavelength range).

  3. ROADM/FOADM properties (connectivity matrix, port wavelength
     restrictions).

  4. Wavelength Converter properties (per network element, may change if
     a common limited shared pool is used).

   In most cases we should be able to combine items (1) and (2) into the
   information in item (3). Except for the number of wavelength
   converters that are available in a shared pool, and the previous
   information is fairly static. In the next two sections we discuss
   dynamic available link bandwidth information.

   5.2.1. Need for Wavelength-Specific Maximum Bandwidth Information

   Difficulties are encountered when trying to use the bandwidth
   accounting methods of [RFC4202] and [RFC3630] to describe the
   availability of wavelengths on a WDM link. The current RFCs give
   three link resource measures: Maximum Bandwidth, Maximum Reservable
   Bandwidth, and Unreserved Bandwidth. Although these can be used to
   describe a WDM span they do not provide the fundamental information
   needed for RWA. We are not given the maximum bandwidth per wavelength
   for the span. If we did then we could use the aforementioned measures
   to tell us the maximum wavelength count and the number of available
   wavelengths.

   For example, suppose we have a 32 channel WDM span, and that the
   system in general supports ITU-T NRZ signals up to NRZ 10Gbps.
   Further suppose that the first 20 channels are carrying 1Gbps
   Ethernet, then the maximum bandwidth would be 320Gbps and the maximum
   reservable bandwidth would be 120Gbps (12 wavelengths).
   Alternatively, consider the case where the first 8 channels are
   carrying 2.5Gbps SDH STM-16 channels, then the maximum bandwidth
   would still be 320Gbps and the maximum reservable bandwidth would be
   240Gbps (24 wavelengths).

   Such information would be useful in the routing with distributed WA
   approach of section 4.3.

   5.2.2. Need for Wavelength-Specific Availability Information

   Even if we know the number of available wavelengths on a link, we
   actually need to know which specific wavelengths are available and
   which are occupied if we are going to run a combined RWA process or
   separate WA process as discussed in section 4.3.  This is currently
   not possible with GMPLS routing extensions.



   Bernstein and Lee Expires August 19, 2008 [Page 22]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   In the routing extensions for GMPLS [RFC4202], requirements for
   layer-specific TE attributes are discussed. The RWA problem for
   optical networks without wavelength converters imposes an additional
   requirement for the lambda (or optical channel) layer: that of
   knowing which specific wavelengths are in use. Note that current
   dense WDM (DWDM) systems range from 16 channels to 128 channels with
   advanced laboratory systems with as many as 300 channels. Given these
   channel limitations and if we take the approach of a global
   wavelength to label mapping or furnishing the local mappings to the
   PCEs then representing the use of wavelengths via a simple bit-map is
   feasible.

   5.2.3. Relationship to Link Bundling and Layering

   When dealing with static DWDM systems, particularly from a SONET/SDH
   or G.709 digital wrapper layer, each lambda looks like a separate
   link. Typically a bunch of unnumbered links, as supported in GMPLS
   routing extensions [RFC4202], would be used to describe a static DWDM
   system. In addition these links can be bundled into a TE link
   ([RFC4202], [RFC4201]) for more efficient dissemination of resource
   information. However, in the case discussed here we want to control a
   dynamic WDM layer and must deal with wavelengths as labels and not
   just as links or component links from the perspective of an upper
   (client) layer. In addition, a typical point to point optical cable
   contains many optical fibers and hence it may be desirable to bundle
   these separate fibers into a TE link. Note that in the no wavelength
   conversion or limited wavelength conversion situations that we will
   need information on wavelength usage on the individual component
   links.

   5.2.4. WSON Routing Information Summary

   The following table summarizes the WSON information that could be
   conveyed via GMPLS routing and attempts to classify that information
   as to its static or dynamic nature and whether that information would
   tend to be associated with either a link or a node.














   Bernstein and Lee Expires August 19, 2008 [Page 23]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


      Information                         Static/Dynamic       Node/Link
      ------------------------------------------------------------------
      Connectivity matrix                 Static               Node
      Per port wavelength restrictions    Static               Node(1)
      WDM link (fiber) lambda ranges      Static               Link
      WDM link channel spacing            Static               Link
      Laser Transmitter range             Static               Link(2)
      Wavelength conversion capabilities  Static(3)            Node
      Maximum bandwidth per Wavelength    Static               Link
      Wavelength Availability             Dynamic(4)           Link

   Notes:

   1. These are the per port wavelength restrictions of an optical
      device such as a ROADM and are independent of any optical
      constraints imposed by a fiber link.

   2. This could also be viewed as a node capability.

   3. This could be dynamic in the case of a limited pool of converters
      where the number available can change with connection
      establishment. Note we may want to include regeneration
      capabilities here since OEO converters are also regenerators.

   4. Not necessarily needed in the case of distributed wavelength
      assignment via signaling.



   While the full complement of the information from the previous table
   is needed in the Combined RWA and the separate Routing and WA
   architectures, in the case of Routing + distribute WA via signaling
   we only need the following information:

      Information                         Static/Dynamic       Node/Link
      ------------------------------------------------------------------
      Connectivity matrix                 Static               Node
      Wavelength conversion capabilities  Static(3)            Node


   Information models and compact encodings for this information is
   provided in [WSON-Info].








   Bernstein and Lee Expires August 19, 2008 [Page 24]


Internet-Draft   Wavelength Switched Optical Networks     February 2008




5.3. Optical Path Computation and Implications for PCE

   As previously noted the RWA problem can be computationally intensive
   [HZang00]. Such computationally intensive path computations and
   optimizations were part of the impetus for the PCE (path computation
   element) architecture.

   As the PCEP defines the procedures necessary to support both
   sequential [PCEP] and global concurrent path computations [PCE-GCO],
   PCE is well positioned to support WSON-enabled RWA computation with
   some protocol enhancement.

   Implications for PCE generally fall into two main categories: (a)
   lightpath constraints and characteristics, (b) computation
   architectures.

   5.3.1. Lightpath Constraints and Characteristics

   For the varying degrees of optimization that may be encountered in a
   network the following models of bulk and sequential lightpath
   requests are encountered:

   o  Batch optimization, multiple lightpaths requested at one time.

   o  Lightpath(s) and backup lightpath(s) requested at one time.

   o  Single lightpath requested at a time.

   PCEP and PCE-GCO can be readily enhanced to support all of the
   potential models of RWA computation.

   Lightpath constraints include:

   o  Bidirectional Assignment of wavelengths

   o  Possible simultaneous assignment of wavelength to primary and
      backup paths.

   o  Tuning range constraint on optical transmitter.

   Lightpath characteristics can include:

   o  Duration information (how long this connection may last)

   o  Timeliness/Urgency information (how quickly is this connection
      needed)


   Bernstein and Lee Expires August 19, 2008 [Page 25]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   5.3.2. Computation Architecture Implications

   When a PCE performs a combined RWA computation per section 4.3. it
   requires accurate an up to date wavelength utilization on all links
   in the network.

   When a PCE is used to perform wavelength assignment (WA) in the
   separate routing WA architecture then the entity requesting WA needs
   to furnish the pre-selected route to the PCE as well as any of the
   lightpath constraints/characteristics previously mentioned. This
   architecture also requires the PCE performing WA to have accurate and
   up to date network wavelength utilization information.

   When a PCE is used to perform routing in a routing with distribute WA
   architecture, then the PCE does not necessarily need the most up to
   date network wavelength utilization information, however timely
   information can contributed to reducing failed signaling attempts
   related to blocking.

   5.3.3. Discovery of RWA Capable PCEs

   The algorithms and network information needed for solving the RWA are
   somewhat specialized and computationally intensive hence not all PCEs
   within a domain would necessarily need or want this capability.
   Hence, it would be useful via the mechanisms being established for
   PCE discovery [RFC5088] to indicate that a PCE has the ability to
   deal with the RWA problem. Reference [RFC5088] indicates that a sub-
   TLV could be allocated for this purpose.

   Recent progress on objective functions in PCE [PCE-OF] would allow
   the operators to flexibly request differing objective functions per
   their need and applications. For instance, this would allow the
   operator to choose an objective function that minimizes the total
   network cost associated with setting up a set of paths concurrently.
   This would also allow operators to choose an objective function that
   results in a most evenly distributed link utilization.

   This implies that PCEP would easily accommodate wavelength selection
   algorithm in its objective function to be able to optimize the path
   computation from the perspective of wavelength assignment if chosen
   by the operators.

5.4. Scaling Implications

   This section provides a summary of the scaling issue for WSON
   routing, signaling and path computation introduced by the concepts
   discussed in this document.



   Bernstein and Lee Expires August 19, 2008 [Page 26]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   5.4.1. Routing

   In large WSONs label availability and cross connect capability
   information being advertised may generate a significant amount of
   routing information.

   5.4.2. Signaling

   When dealing with a large number of simultaneous end-to-end
   wavelength service requests and service deletions the network may
   have to process a significant number of forward and backward service
   messages. Also, similar situation possibly happens in the case of
   link or node failure, if the WSON support dynamic restoration
   capability.

   5.4.3. Path computation

   If a PCE is handling path computation requests for end-to-end
   wavelength services within the WSON, then the complexity of the
   network and number of service path computation requests being sent to
   the PCE may have an impact on the PCEs ability to process requests in
   a timely manner.

6. Security Considerations

   This document has no requirement for a change to the security models
   within GMPLS and associated protocols. That is the OSPF-TE, RSVP-TE,
   and PCEP security models could be operated unchanged.

   However satisfying the requirements for RWA using the existing
   protocols may significantly affect the loading of those protocols.
   This makes the operation of the network more vulnerable to denial of
   service attacks. Therefore additional care maybe required to ensure
   that the protocols are secure in the WSON environment.

   Furthermore the additional information distributed in order to
   address the RWA problem represents a disclosure of network
   capabilities that an operator may wish to keep private. Consideration
   should be given to securing this information.

7. IANA Considerations

   This document makes no request for IANA actions.

8. Acknowledgments

   The authors would like to thank Adrian Farrel for many helpful
   comments that greatly improved the contents of this draft.


   Bernstein and Lee Expires August 19, 2008 [Page 27]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   This document was prepared using 2-Word-v2.0.template.dot.

















































   Bernstein and Lee Expires August 19, 2008 [Page 28]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


9. References

9.1. Normative References

   [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
             Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC3471] Berger, L., "Generalized Multi-Protocol Label Switching
             (GMPLS) Signaling Functional Description", RFC 3471,
             January 2003.

   [RFC3630] Katz, D., Kompella, K., and D. Yeung, "Traffic Engineering
             (TE) Extensions to OSPF Version 2", RFC 3630, September
             2003.

   [RFC3945] Mannie, E., "Generalized Multi-Protocol Label Switching
             (GMPLS) Architecture", RFC 3945, October 2004.

   [RFC4201] Kompella, K., Rekhter, Y., and L. Berger, "Link Bundling in
             MPLS Traffic Engineering (TE)", RFC 4201, October 2005.

   [RFC4202] Kompella, K. and Y. Rekhter, "Routing Extensions in Support
             of Generalized Multi-Protocol Label Switching (GMPLS)", RFC
             4202, October 2005.

   [RFC4203] Kompella, K. and Y. Rekhter, "OSPF Extensions in Support of
             Generalized Multi-Protocol Label Switching (GMPLS)", RFC
             4203, October 2005.

   [RFC4328] Papadimitriou, D., "Generalized Multi-Protocol Label
             Switching (GMPLS) Signaling Extensions for G.709 Optical
             Transport Networks Control", RFC 4328, January 2006.

   [G.694.1] ITU-T Recommendation G.694.1, "Spectral grids for WDM
             applications: DWDM frequency grid", June, 2002.

   [RFC5088] J.L. Le Roux, J.P. Vasseur, Yuichi Ikejiri, and Raymond
             Zhang, "OSPF protocol extensions for Path Computation
             Element (PCE) Discovery", January 2008.

   [PCE-GCO] Y. Lee, J.L. Le Roux, D. King, and E. Oki, "Path
             Computation Element Communication Protocol (PCECP)
             Requirements and Protocol Extensions In Support of Global
             Concurrent Optimization", work in progress, draft-ietf-pce-
             global-concurrent-optimization-01.txt, November 2007.





   Bernstein and Lee Expires August 19, 2008 [Page 29]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   [PCEP]    J.P. Vasseur and J.L. Le Roux (Editors), "Path Computation
             Element (PCE) Communication Protocol (PCEP)", work in
             progress, draft-ietf-pce-pcep-10.txt, February 2008.

   [PCE-OF] J.L. Le Roux, J.P. Vasseur, and Y. Lee, "Encoding of
             Objective Functions in Path Computation Element (PCE)
             communication and discovery protocols", work in progress,
             draft-ietf-pce-of-01.txt, February 2008.

   [WSON-Info] G. Bernstein, Y. Lee, D. Li, W. Imajuku," Routing and
             Wavelength Assignment Information for Wavelength Switched
             Optical Networks", draft-bernstein-ccamp-wson-info-02.txt,
             February, 2008.

9.2. Informative References

   [HZang00] H. Zang, J. Jue and B. Mukherjeee, "A review of routing and
             wavelength assignment approaches for wavelength-routed
             optical WDM networks", Optical Networks Magazine, January
             2000.

   [Coldren04]    Larry A. Coldren, G. A. Fish, Y. Akulova, J. S.
             Barton, L. Johansson and C. W. Coldren, "Tunable
             Seiconductor Lasers: A Tutorial", Journal of Lightwave
             Technology, vol. 22, no. 1, pp. 193-202, January 2004.

   [Chu03]   Xiaowen Chu, Bo Li and Chlamtac I, "Wavelength converter
             placement under different RWA algorithms in wavelength-
             routed all-optical networks", IEEE Transactions on
             Communications, vol. 51, no. 4, pp. 607-617, April 2003.

   [Buus06]    Jens Buus EJM, "Tunable Lasers in Optical Networks",
             Journal of Lightware Technology, vol. 24, no. 1, pp. 5-11,
             January 2006.

   [Basch06] E. Bert Bash, Roman Egorov, Steven Gringeri and Stuart
             Elby, "Architectural Tradeoffs for Reconfigurable Dense
             Wavelength-Division Multiplexing Systems", IEEE Journal of
             Selected Topics in Quantum Electronics, vol. 12, no. 4, pp.
             615-626, July/August 2006.

   [Otani]  T. Otani, H. Guo, K. Miyazaki, D. Caviglia, "Generalized
             Labels of Lambda-Switching Capable Label Switching Routers
             (LSR)", work in progress: draft-otani-ccamp-gmpls-lambda-
             labels-01.txt, November 2007.





   Bernstein and Lee Expires August 19, 2008 [Page 30]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   [Winzer06]    Peter J. Winzer and Rene-Jean Essiambre, "Advanced
             Optical Modulation Formats", Proceedings of the IEEE, vol.
             94, no. 5, pp. 952-985, May 2006.

   [G.652] ITU-T Recommendation G.652, Characteristics of a single-mode
             optical fibre and cable, June 2005.

   [G.653] ITU-T Recommendation G.653, Characteristics of a dispersion-
             shifted single-mode optical fibre and cable, December 2006.

   [G.654] ITU-T Recommendation G.654, Characteristics of a cut-off
             shifted single-mode optical fibre and cable, December 2006.

   [G.655] ITU-T Recommendation G.655, Characteristics of a non-zero
             dispersion-shifted single-mode optical fibre and cable,
             March 2006.

   [G.656] ITU-T Recommendation G.656, Characteristics of a fibre and
             cable with non-zero dispersion for wideband optical
             transport, December 2006.

   [G.959.1] ITU-T Recommendation G.959.1, Optical Transport Network
             Physical Layer Interfaces, March 2006.

   [G.694.1] ITU-T Recommendation G.694.1, Spectral grids for WDM
             applications: DWDM frequency grid, June 2002.

   [G.694.2] ITU-T Recommendation G.694.2, Spectral grids for WDM
             applications: CWDM wavelength grid, December 2003.

   [G.Sup39] ITU-T Series G Supplement 39, Optical system design and
             engineering considerations, February 2006.

   [G.Sup43] ITU-T Series G Supplement 43, Transport of IEEE 10G base-R
             in optical transport networks (OTN), November 2006.

   [Imajuku] W. Imajuku, Y. Sone, I. Nishioka, S. Seno, "Routing
             Extensions to Support Network Elements with Switching
             Constraint", work in progress: draft-imajuku-ccamp-rtg-
             switching-constraint-02.txt, July 2007.

   [RFC4054] Strand, J. and  A. Chiu, "Impairments and Other Constraints
             on Optical Layer Routing", RFC 4054, May 2005.

   [RFC4606] Mannie, E. and D. Papadimitriou, "Generalized Multi-
             Protocol Label Switching (GMPLS) Extensions for Synchronous
             Optical Network (SONET) and Synchronous Digital Hierarchy
             (SDH) Control", RFC 4606, August 2006.


   Bernstein and Lee Expires August 19, 2008 [Page 31]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


10. Contributors

   Snigdho Bardalai
   Fujitsu
   Email: Snigdho.Bardalai@us.fujitsu.com

   Diego Caviglia
   Ericsson
   Via A. Negrone 1/A 16153
   Genoa Italy

   Phone: +39 010 600 3736
   Email: diego.caviglia@(marconi.com, ericsson.com)

   Daniel King
   Aria Networks
   Email: daniel.king@aria-networks.com

   Itaru Nishioka
   NEC Corp.
   1753 Simonumabe, Nakahara-ku, Kawasaki, Kanagawa 211-8666
   Japan
   Phone: +81 44 396 3287
   Email: i-nishioka@cb.jp.nec.com

   Lyndon Ong
   Ciena
   Email: Lyong@Ciena.com

   Jonathan Sadler
   Tellabs
   Email: Jonathan.Sadler@tellabs.com

Author's Addresses

   Greg Bernstein (ed.)
   Grotto Networking
   Fremont, CA, USA

   Phone: (510) 573-2237
   Email: gregb@grotto-networking.com









   Bernstein and Lee Expires August 19, 2008 [Page 32]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


   Young Lee (ed.)
   Huawei Technologies
   1700 Alma Drive, Suite 100
   Plano, TX 75075
   USA

   Phone: (972) 509-5599 (x2240)
   Email: ylee@huawei.com


   Wataru Imajuku
   NTT Network Innovation Labs
   1-1 Hikari-no-oka, Yokosuka, Kanagawa
   Japan

   Phone: +81-(46) 859-4315
   Email: imajuku.wataru@lab.ntt.co.jp




Intellectual Property Statement

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at
   http://www.ietf.org/ipr.

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at
   ietf-ipr@ietf.org.






   Bernstein and Lee Expires August 19, 2008 [Page 33]


Internet-Draft   Wavelength Switched Optical Networks     February 2008


Disclaimer of Validity

   This document and the information contained herein are provided on an
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND
   THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS
   OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF
   THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
   WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Copyright Statement

   Copyright (C) The IETF Trust (2008).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.

Acknowledgment

   Funding for the RFC Editor function is currently provided by the
   Internet Society.




























   Bernstein and Lee Expires August 19, 2008 [Page 34]