| Network Working Group | O. Gonzalez de Dios, Ed. |
| Internet-Draft | Telefónica I+D |
| Intended status: Standards Track | R. Casellas, Ed. |
| Expires: October 13, 2013 | CTTC |
| F. Zhang | |
| Huawei | |
| X. Fu | |
| ZTE | |
| D. Ceccarelli | |
| Ericsson | |
| I. Hussain | |
| Infinera | |
| April 11, 2013 |
Framework and Requirements for GMPLS based control of Flexi-grid DWDM networks
draft-ogrcetal-ccamp-flexi-grid-fwk-02
This document defines a framework and the associated control plane requirements for the GMPLS based control of flexi-grid DWDM networks. To allow efficient allocation of optical spectral bandwidth for high bit-rate systems, the International Telecommunication Union Telecommunication Standardization Sector (ITU-T) has extended the recommendations [G.694.1] and [G.872] to include the concept of flexible grid: a new DWDM grid has been developed within the ITU-T Study Group 15 by defining a set of nominal central frequencies, smaller channel spacings and the concept of "frequency slot". In such environment, a data plane connection is switched based on allocated, variable-sized frequency ranges within the optical spectrum.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
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."
This Internet-Draft will expire on October 13, 2013.
Copyright (c) 2013 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
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 [RFC2119].
The term "Flexible grid" (flexi-grid for short) as defined by the International Telecommunication Union Telecommunication Standardization Sector (ITU-T) Study Group 15 in the latest version of [G.694.1], refers to the updated set of nominal central frequencies (a frequency grid), channel spacings and optical spectrum management/allocation considerations that have been defined in order to allow an efficient and flexible allocation and configuration of optical spectral bandwidth for high bit-rate systems.
A key concept of flexi-grid is the "frequency slot"; a variable-sized optical frequency range that can be allocated to a data connection. As detailed later in the document, a frequency slot is characterized by its nominal central frequency and its slot width which, as per [G.694.1], is constrained to be a multiple of a given slot width granularity.
Compared to a traditional fixed grid network, which uses fixed size optical spectrum frequency ranges or "frequency slots" with typical channel separations of 50 GHz, a flexible grid network can select its media channels with a more flexible choice of slot widths, allocating as much optical spectrum as required, allowing high bit rate signals (e.g., 400G, 1T or higher) that do not fit in the fixed grid.
From a networking perspective, a flexible grid network is assumed to be a layered network [G.872][G.800] in which the flexi-grid layer (also referred to as the media layer) is the server layer and the OCh Layer (also referred to as the signal layer) is the client layer. In the media layer, switching is based on a frequency slot, and the size of a media channel is given by the properties of the associated frequency slot. In this layered network, the media channel itself can be dimensioned to contain one or more Optical Channels.
As described in [RFC3945], GMPLS extends MPLS from supporting only Packet Switching Capable (PSC) interfaces and switching to also support four new classes of interfaces and switching that include Lambda Switch Capable (LSC).
A Wavelength Switched Optical Network (WSON), addressed in [RFC6163], is a term commonly used to refer to the application/deployment of a Generalized Multi-Protocol Label Switching (GMPLS)-based control plane for the control (provisioning/recovery, etc) of a fixed grid WDM network. [editors' note: we need to think of the relationship of WSON and OCh switching. Are they equivalent? WSON includes regeneration, OCh does not? decoupling of lambda/OCh/OCC]
This document defines the framework for a GMPLS-based control of flexi-grid enabled DWDM networks (in the scope defined by ITU-T layered Optical Transport Networks [G.872], as well as a set of associated control plane requirements. An important design consideration relates to the decoupling of the management of the optical spectrum resource and the client signals to be transported. [Editor's note: a point was raised during the meeting that WSON has not made the separation between Och and Lambda (spectrum and signal are bundled). The document will consider and evaluate the relationship later].
[Editors' note: this document will track changes and evolutions of [G.694.1] [G.872] documents until their final publication. This document is not expected to become RFC until then.]
[Editor's note: -00 as agreed during IETF83, the consideration of the concepts of Super-channel (a collection of one or more frequency slots to be treated as unified entity for management and control plane) and consequently Contiguous Spectrum Super-channel (a super-channel with a single frequency slot) and Split-Spectrum super-channel (a super-channel with multiple frequency slots) is postponed until the ITU-T data plane includes such physical layer entities, e.g., an ITU-T contribution exists. ITU-T is still discussing B100G Architecture]
[Editors' note: -01 this version reflects the agreements made during IETF84, notably concerning the focus in the media layer, terminology updates post ITU-T September meeting in Geneva and the deprecation of the ROADM term, in favor of the more concrete media layer switching element (media channel matrix).]
[Editors' note: -01 in partial answer to Gert question on the layered model, [G.872] footnote explains that this separation is necessary to allow the description of media elements that may act on more than a single OCh-P signal. See appendix IV within.]
FS: Frequency Slot
NCF: Nominal Central Frequency
OCh: Optical Channel
OCh-P: Optical Channel Payload
OCh-O: Optical Channel Overhead
OCC: Optical Channel Carrier
SWG: Slot Width Granularity
The following is a list of terms (see [G.694.1] and [G.872]) reproduced here for completeness. [Editors' note: regarding wavebands, we agreed NOT to use the term in flexi-grid. The term has been used inconsistently in fixed-grid networks and overlaps with the definition of frequency slot. If need be, a question will be sent to ITU-T asking for clarification regarding wavebands.]
Where appropriate, this documents also uses terminology and lexicography from [RFC4397].
[Editors' note: *important* these terms are not yet final and they may change / be replaced or obsoleted at any time.]
-5 -4 -3 -2 -1 0 1 2 3 4 5 <- values of n
...+--+--+--+--+--+--+--+--+--+--+-
^
193.1 THz <- anchor frequency
Figure 1: Anchor frequency and set of nominal central frequencies
Frequency Slot 1 Frequency Slot 2
------------- -------------------
| | | |
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11
..--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--...
------------- -------------------
^ ^
Central F = 193.1THz Central F = 193.14375 THz
Slot width = 25 GHz Slot width = 37.5 GHz
Figure 2: Example Frequency slots
Frequency Slot 1
-------------
| |
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11
..--+--+--+--+--X--+--+--+--+--+--+--+--+--+--+--+--...
Frequency Slot 2
-------------------
| |
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11
..--+--+--+--+--X--+--+--+--+--+--+--+--+--+--+--+--...
===============================================
Effective Frequency Slot
-------------
| |
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11
..--+--+--+--+--X--+--+--+--+--+--+--+--+--+--+--+--...
Figure 3: Effective Frequency Slot
Frequency Slot 1
-------------
| |
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11
..--+--+--X--+--+--+--+--+--+--+--+--+--+--+--+--+--...
Frequency Slot 2
-------------------
| |
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11
..--+--+--+--+--X--+--+--+--+--+--+--+--+--+--+--+--...
===============================================
Invalid Effective Frequency Slot - (n, m?)
----------
| |
-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11
..--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--+--...
Figure 4: Invalid Effective Frequency Slot
Note that by definition a network media channel only supports a single OCh-P network connection, but the architecture is flexible enough to support the case where a single Optical Channel Payload (OCh-P) is transported over multiple (N) network media channels (see Figure 5) which are non-necessarily adjacent (e.g., there may not exist an equivalent network media channel that can be represented as the union of media channels with a single, valid equivalent frequency slot). Whether a single OCh-P is transported over just one or more network media channels is an aspect that corresponds to the mapping of the signal layer to the media layer.
OCh-P Signal Layer
+--------X----------+
/ \ Media Layer
/ \
Network Media Channel #1 ... Network Media Channel #N
+------------o-----------+ +----------X-----------+
| | | |
... +---+---+---+---+---+---+---+---+---+---+---+--+---+---+---+---+---...
Figure 5: A single OCh-P transported over multiple network media channels
[Editors' note: we are not discarding O/E/O. If defined in a ITU-T network reference model with trail/terminations, considering optical channels i.e. with well-defined interfaces, reference points, and architectures. The implications of O/E/O will be also addressed once we have another context that includes them. In OTN from an OCh point of view end to end means from transponder to transponder, so if there is a 3R from ingress to egress there are 2 OCh which can have different 'n' and 'm'].
Media Channel Frequency Slot
+-------------------------------X------------------------------+
| |
| OCh-P Frequency Slot OCh-P Frequency Slot |
| +------------X-----------+ +----------X-----------+ |
| | OCh-P | | OCh-P | |
| | o | | o | |
| | | | | | | |
-4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12
... +---+---+---+---+---+---+---+---+---+---+---+--+---+---+---+---+---...
... <- Network Media Channel-> <- Network Media Channel->
... <------------------------ Media Channel ----------------------->
X - Frequency Slot Central Frequency
o - signal central frequency
Figure 6: Example of Media Channel / Network Media Channels and associated frequency slots
In summary, the concept of frequency slot is a logical abstraction that represents a frequency range while the media layer represents the underlying media support. Media Channels are media associations, characterized by their (effective) frequency slot, respectively; and media channels are switched in media channel matrixes. In Figure 6 , a Media Channel has been configured and dimensioned to support two OCh-P, each transported in its own OCh-P frequency slot.
The following terms are defined in the scope of a GMPLS control plane.
Similar to fixed grid networks, a flexible grid network is also constructed from subsystems that include Wavelength Division Multiplexing (WDM) links, tunable transmitters and receivers, i.e, media elements including media layer switching elements (media matrices), as well as electro-optical network elements, all of them with flexible grid characteristics.
[Editors' Note: In the scope of this document, and despite is informal use, the term Reconfigurable Optical Add / Drop Multiplexer, (ROADM) is avoided, in favor on media matrix. This avoid ambiguity. A ROADM can be implemented in terms on media matrices. Informationally, this document may provide an appendix on possible implementations of flexi-ROADMs in terms of media layer switching elements or matrices. XF: Whether ROADM is used or not doesn't matter with GMPLS Control Plane. I suggest to delete this statement. We may check G.798. Likewise, modeling of filters is out of scope of the current document IETF84, and is also considered implementation specific.]
As stated in [G.694.1] the flexible DWDM grid defined in Clause 7 has a nominal central frequency granularity of 6.25 GHz and a slot width granularity of 12.5 GHz. However, devices or applications that make use of the flexible grid may not be capable of supporting every possible slot width or position. In other words, applications may be defined where only a subset of the possible slot widths and positions are required to be supported. For example, an application could be defined where the nominal central frequency granularity is 12.5 GHz (by only requiring values of n that are even) and that only requires slot widths as a multiple of 25 GHz (by only requiring values of m that are even).
[TODO: section needs to be rewritten, remove redundancy].
Optical transmitters/receivers may have different tunability constraints, and media channel matrixes may have switching restrictions. Additionally, a key feature of their implementation is their highly asymmetric switching capability which is described in [RFC6163] in detail. Media matrices include line side ports which are connected to DWDM links and tributary side input/output ports which can be connected to transmitters/receivers.
A set of common constraints can be defined:
[Editor's note: different configurations such as C/CD/CDC will be added later. This section should state specifics to media channel matrices, ROADM models need to be moved to an appendix].
OCh AP Trail (OCh) OCh AP
O- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - O
| |
--- OCh-P OCh-P ---
\ / source sink \ /
+ +
| OCh-P OCh-P Network Connection OCh-P |
O TCP - - - - - - - - - - - - - - - - - - - - - - - - - - - - -TCP O
| |
|Channel Port Network Media Channel Channel Port |
O - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - O
| |
+--------+ +-----------+ +----------+
| \ (1) | OCh-P LC | (1) | OCh-P LC | (1) / |
| \----|-----------------|-----------|-------------------|-------/ |
+--------+ Link Channel +-----------+ Link Channel +----------+
Media Channel Media Channel Media Channel
Matrix Matrix Matrix
(1) - Matrix Channel
Figure 7: Layered Network Model according to G.805
This section presents an overview of the OCh / flexi-grid layered network model defined by ITU-T. [Editors' note: OTN hierarchy is not fully covered. It is important to understand, where the FSC sits in the OTN hierarchy. This is also important from control plane perspective as this layer becomes the connection end points of optical layer service].
[Editors' note: we are replicating the figure here for reference, until the ITU-T document is official.
The goal of this section is to provide an insight of the application of GMPLS to flexi-grid networks, while specific requirements are covered in the next section. The present framework is aimed at controlling the so called media layer within the OTN hierarchy. Specifically, the GMPLS control of the media layer deals with the establishment of media channels, which are switched in media channel matrixes. GMPLS labels locally represent the media channel and its associated frequency slot. [Editors'note: As agreed during IETF84, current focus is on the media layer.]
Also, this sections provides a mapping of the mapping ITU-T G.872 architectural aspects to GMPLS/Control plane terms, and see the relationship between the architectural concept/construct of media channel and its control plane representations (e.g. as a TE link).
From a theoretical / abstract point of view, a fiber can be modeled has having a frequency slot that ranges from (-inf, +inf). This representation helps understand the relationship between frequency slots / ranges.
The frequency slot is a local concept that applies locally to a component / element. When applied to a media channel, we are referring to its effective frequency slot as defined in [G.872].
The association of a filter, a fiber and a filter is a media channel in its most basic form, which from the control plane perspective may modeled as a (physical) TE-link with a contiguous optical spectrum at start of day. A means to represent this is that the portion of spectrum available at time t0 depends on which filters are placed at the ends of the fiber and how they have been configured. Once filters are placed we have the one hop media channel. In practical terms, associating a fiber with the terminating filters determines the usable optical spetrum.
-----------------+ +-----------------+
| |
+--------+ +--------+
| | | | +---------
---o| ================================== o--|
| | Fiber | | | --\ /--
---o| | | o--| \/
| | | | | /\
---o| ================================== o--| --/ \--
| Filter | | Filter | |
| | | | +---------
+--------+ +--------+
| |
|---------- Basic Media Channel ---------|
-----------------+ +-----------------+
--------+ +--------+
|-----------------------------------------|
LSR | TE link | LSR
|-----------------------------------------|
+--------+ +--------+
Figure 8: (Basic) Media channel and TE link
Additionally, when a cross-connect for a specific frequency slot is considered, the underlying media support is still a media channel, augmented, so to speak, with a bigger association of media elements and a resulting effective slot. When this media channel is the result of the association of basic media channels and media layer matrix cross-connects, this architectural construct can be represented as / corresponds to a Label Switched Path (LSP) from a control plane perspective. In other words, It is possible to "concatenate" several media channels (e.g. Patch on intermediate nodes) to create a single media channel.
------------+ +--------------------------------+ +----------
| | | |
+------+ +------+ +------+ +------+
| | | | +----------+ | | | |
---o| ========= o---| |---o ========= o--
| | Fiber | | | --\ /-- | | | Fiber | |
---o| | | o---| \/ |---o | | o--
| | | | | /\ | | | | |
---o| ========= o---***********|---o ========= o--
|Filter| |Filter| | | |Filter| |Filter|
| | | | | | | |
+------+ +------+ +------+ +------+
| | | |
<- Basic Media -> <- Matrix -> <- Basic Media->
|Channel| Channel |Channel|
------------+ +--------------------------------+ +----------
<--------------------- Media Channel ----------------->
-----+ +---------------+ +-------
|-------------------| |-------------------|
LSR | TE link | LSR | TE link | LSR
|-------------------| |-------------------|
-----+ +---------------+ +-------
Figure 9: Extended Media Channel
Additionally, if appropriate, it can also be represented as a TE link or Forwarding Adjacency (FA), augmenting the control plane network model.
-----------+ +--------------------------------+ +-----------
| | | |
+------+ +------+ +------+ +------+
| | | | +----------+ | | | |
--o| ========= o---| |---o ========= o---
| | Fiber | | | --\ /-- | | | Fiber | |
--o| | | o---| \/ |---o | | o---
| | | | | /\ | | | | |
--o| ========= o---***********|---o ========= o---
|Filter| |Filter| | | |Filter| |Filter|
| | | | | | | |
+------+ +------+ +------+ +------+
| | | |
-----------+ +--------------------------------+ +-----------
<--------------------------- Media Channel ----------->
+-----+ +------
|--------------------------------------------------------|
LSR | TE link | LSR
|--------------------------------------------------------|
+-----+ +------
Figure 10: Extended Media Channel / TE Link / FA
The flexi-grid LSP is seen as a control plane representation of a media channel. Since network media channels are media channels, an LSP may also be the control plane representation of a network media channel, in a particular context. From a control plane perspective, the main difference (regardless of the actual effective frequency slot which may be dimensioned arbitrarily) is that the LSP that represents a network media channel also includes the endpoints (transceivers) , including the cross-connects at the ingress / egress nodes. The ports towards the client can still be represented as interfaces from the control plane perspective.
Figure 11 describes an LSP routed along 3 nodes. The LSP is terminated before the optical matrix of the ingress and egress nodes and can represent a Media Channel. This case does NOT (and cannot) represent a network media channel as it does not include (and cannot include) the transceivers.
----------+ +------------------------------------+ +--------------
| | | |
+------+ +--------+ +--------+ +------+
| | | | +----------+ | | | |
-o| ========= o---| |---o ========= o---
| | Fiber | | | --\ /-- | | | Fiber | |
-o|>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>o---
| | | | | /\ | | | | |
-o| ========= o---***********|---o ========= o---
|Filter| | Filter | | | | Filter | |Filter|
| | | | | | | |
+------+ +--------+ +--------+ +------+
| | | |
----------+ +------------------------------------+ +--------------
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> LSP >>>>>>>>>>>>>>>>>>>>>>>>
-----+ +---------------+ +-----
|----------------------| |--------------------|
LSR | TE link | LSR | TE link | LSR
|----------------------| |--------------------|
-----+ +---------------+ +-----
Figure 11: Flex-grid LSP representing a media channel that starts at the filter of the outgoing interface of the ingress LSR and ends at the filter of the incoming interface of the egress LSR
In Figure 12 a Network Media Channel is represented as terminated at the DWDM side of the transponder, this is commonly named as OCh-trail connection.
|-------------------------- Network Media Channel ---------------------------|
+------------------------+ +------------------------+
| | | |
+------+ +------+ +------+ +------+
| | +------+ | | | | +------+ | |OCh-P
OCh-P| o-| |-o | +-----+ | o-| |-o |sink
src | | | | | =====+-+ +-+======| | | | | O---|R
T|***o******o******************************************************************
| | | \ / | | | | | | | | | \ / | | |
| o-| \/ |-o =====| | | |======| o-| \/ |-o |
| | | /\ | | | +-+ +-+ | | | /\ | | |
| o-| / \ |-o | | \/ | | o-| / \ |-o |
|Filter| | | |Filter| | /\ | |Filter| | | |Filter|
+------+ | | +------+ +------+ +------+ | | +------+
| | | | | | | |
+------------------------+ +------------------------+
LSP
<---------------------------------------------------------------------------->
LSP
<-------------------------------------------------------------------------->
+-----+ +--------+ +-----+
o--- | |-----------------------| |--------------------| |---o
| LSR | TE link | LSR | TE link | LSR |
| |-----------------------| |--------------------| |
+-----+ +--------+ +-----+
Figure 12: LSP representing a network media channel (OCh-Trail)
In a third case, a Network Media Channel terminated on the Filter ports of the Ingress and Egress nodes. This is named in G.872 as OCh-NC (we need to discuss the implications, if any, once modeled at the control plane level of models B and C).
|----------------------- Network Media Channel --------------------|
+------------------------+ +------------------------+
+------+ +------+ +------+ +------+
| | +-----+ | | | | +------+ | |
| o-| |-o | +------+ | o-| |-o |
| | | | | =====+-+ +-+=====| | | | | O---|R
T|--o******o*************************************************************
| | | \ / | | | | | | | | | \ / | | |
| o-| \/ |-o =====| | | |=====| o-| \/ |-o |
| | | /\ | | | +-+ +-+ | | | /\ | | |
| o-| / \ |-o | | \/ | | o-| / \ |-o |
|Filter| | | |Filter| | /\ | |Filter| | | |Filter|
+------+ | | +------+ +------+ +------+ | | +------+
| | | | | | | |
+------------------------+ +------------------------+
<------------------------------------------------------------------->
LSP
LSP
<----------------------------------------------------------------->
+-----+ +--------+ +-----+
o---| |----------------------| |----------------------| |---o
| LSR | TE link | LSR | TE link | LSR |
| |----------------------| |----------------------| |
+-----+ +--------+ +-----+
Figure 13: LSP representing a network media channel (OCh-P NC)
[Note: not clear the difference, from a control plane perspective, of figs Figure 12 and Figure 13.]
+--------------+ +--------------+
| OCh-P | TE | OCh-P | Virtual TE
| | link | | link
| Matrix |o- - - - - - - - - - - - - o| Matrix |o- - - - - -
+--------------+ +--------------+
| +---------+ |
| | Media | |
|o------| Channel |---------o|
| |
| Matrix |
+---------+
Figure 14: MRN/MLN topology view with TE link / FA
Applying the notion of hierarchy at the media layer, by using the LSP as a FA, the media channel created can support multiple (sub) media channels. [Editot note : a specific behavior related to Hierarchies will be verified at a later point in time].
Note that there is only one media layer switch matrix (one implementation is FlexGrid ROADM) in SSON, while "signal layer LSP is mainly for the purpose of management and control of individual optical signal". Signal layer LSPs (OChs) with the same attributions (such as source and destination) could be grouped into one media-layer LSP (media channel), which has advantages in spectral efficiency (reduce guard band between adjacent OChs in one FSC) and LSP management. However, assuming some network elements indeed perform signal layer switch in SSON, there must be enough guard band between adjacent OChs in one media channel, in order to compensate filter concatenation effect and other effects caused by signal layer switching elements. In such condition, the separation of signal layer from media layer cannot bring any benefit in spectral efficiency and in other aspects, but make the network switch and control more complex. If two OChs must switch to different ports, it is better to carry them by diferent FSCs and the media layer switch is enough in this scenario.
[Editor's note: The considered topology view is a layered network, in which the media layer corresponds to the server layer (flexigrid) and the signal layer corresponds to the client layer (Och). This data plane modeling considers the flexigrid and the OCh as separate layers, However, this has implications on the interop/interworking with WSON and OCh switching. We need to manage a MRN for OCh and stitching for WSON? In other words, a key part of the fwk is to define how can we have MRN/MLN hierarchical relationship with Och/FS and yet stitching 1:1 between WSON and SSON? In this line: how does OCh switching and WSON relate, actually?]
[Editor's note: formal requirements such as noted in the comments will be added in a later version of the document].
The data plane model decouples media and signal. The data plane definition relative to the the signal mapping are still ongoing. Until they are closed, the Control Plane considered is based on a MRN/MLN network in which the Control Plane differentiates signal LSPs and media LSPs. This will be revised when the data plane definition are finalized.
The signal does impose restriction on the flexi-LS by constraining the effective frequency slot. The central frequency of each hop should be same along end-to-end media or signal LSP because of Spectrum Continuity Constraint. [Editors Note: the effective frequency slot should fit, which clause mandate that the central frequency is the same ??] Otherwise some nodes need to conver the central frequency along media or signal LSP.
[Editors' note: preliminary text based on IETF85 side-meeting minutes]
A media channel has an associated effective frequency slot. From the perspective of network control and management, this effective slot is seen as the "usable" frequency slot end to end. The establishment of an LSP related the establishment of the media channel and effective frequency slot.
In this context, when used unqualified, the frequency slot is a local term, which applies at each hop. An effective frequency slot applies at the media chall (LSP) level
A "service" request is characterized as a minimum, by its required effective slot width. This does not preclude that the request may add additional constraints such as imposing also the nominal central frequency. A given frequency slot is requested for the media channel say, with the Path message. Regardless of the actual encoding, the Path message sender descriptor sender_tspec shall specify a minimum frequency slot width that needs to be fulfilled.
In order to allocate a proper effective frequency slot for a LSP, the signaling should specify its required slot width.
An effective frequency slot must equally be described in terms of a central nominal frequency and its slot width (in terms of usable spectrum of the effective frequency slot). That is, one must be able to obtain an end-to-end equivalent n and m parameters. We refer to this as the "effective frequency slot of the media channel/LSP must be valid".
In GMPLS the requested effective frequency slot is represented to the TSpec and the effective frequency slot is mapped to the FlowSpec.
The switched element corresponds in GMPLS to the 'label'. As in flexi-grid the switched element is a frequency slot, the label represents a frequency slot. Consequently, the label in flexi-grid must convey the necessary information to obtain the frequency slot characteristics (i.e, center and width, the n and m parameters). The frequency slot is locally identified by the label
The local frequency slot may change at each hop, typically given hardware constraints (e.g. a given node cannot support the finest granularity). Locally n and m may change. As long as a given downstream node allocates enough optical spectrum, m can be different along the path. This covers the issue where concrete media matrices can have different slot width granularities. Such "local" m will appear in the allocated label that encodes the frequency slot as well as the flow descriptor flowspec.
Different modes are considered: RSA with explicit label control , and for R+DSA, the GMPLS signaling procedure is similar to the one described in section 4.1.3 of [RFC6163] except that the label set should specify the available nominal central frequencies that meet the slot width requirement of the LSP. The intermediate nodes can collect the acceptable central frequencies that meet the slot width requirement hop by hop. The tail-end node also needs to know the slot width of a LSP to assign the proper frequency resource. Compared with [RFC6163], except identifying the resource (i.e., fixed wavelength for WSON and frequency resource for flexible grids), the other signaling requirements (e.g., unidirectional or bidirectional, with or without converters) are the same as WSON described in the section 6.1 of [RFC6163].
Regarding how a GMPLS control plane can assign n and m, different cases can apply:
In consequence, an entity such as a PCE can make sure that the n and m stay the same along the path. Any constraint (including frequency slot and width granularities) is taken into account during path computation. alternatively, A PCE (or a source node) can compute a path and the actual frequency slot assignment is done, for example, with a distributed (signaling) procedure:
For Figure Figure 15 the effective slot is valid by ensuring that the minimum m is greater than the requested m. The effective slot (intersection) is the lowest m (bottleneck).
|Path(m_req) | ^ |
|---------> | # |
| | # ^
-^----------------^------------------#-----------------#--
Effective # # # #
FS n, m # . . . . . . . .#. . . . . . . . . # . . . . . .. . .# <-fixed
# # # # n
-v----------------v------------------#-----------------#---
| | # v
| | # Resv |
| | v <------ |
| | | flowspec(n, m_a)|
| | <--------| |
| | flowspec (n, |
<--------| min(m_a, m_b))
flowspec (n, |
min(m_a, m_b, m_c))
Figure 15: Distributed allocation with different m and same n
|Path(m_req) ^ |
|---------> # | |
| # ^ ^
-^----------------#------------------#-----------------#----------
Effective # # # #
FS n, m # # # #
# # # #
-v----------------v------------------#-----------------#----------
| | # v
| | # Resv |
| | v <------ |
| | |flowspec(n_a, m_a)
| | <--------| |
| | flowspec (FSb [intersect] FSa)
<--------|
flowspec ([intersect] FSa,FSb,FSc)
Figure 16: Distributed allocation with different m and different n
For Figure Figure 16 the effective slot is valid by ensuring that it is valid at each hop in the upstream direction. The intersection needs to be computed. Invalid slots could result otherwise.
Note, when a media channel is bound to one OCh-P (i.e is a Network media channel), the EFS must be the one of the Och-P. The media channel setup by the LSP may contains the EFS of the network media channel EFS. This is an endpoint property, the egress and ingress SHOULD constrain the EFS to Och-P EFS [q: can't the endpoint be configured based on the final allocation? constraint or output?].
0 1 2 3 4 ...
|<-------2m---------->|
|--------na-----------|
---------X---------------X------------
|----------nb--------|
|<-------2m--------->|
------------------------------------
|<2mreq>|
(na + m) - (nb - m) >= 2mreq where nb >= na
(nb - na) <= 2(m - mreq)
Figure 17: Valid frequency slot case d)
Also note that the case d) applies for one cannot assume that m = m_requested. If m = m_requested, one cannot ensure that the effective slot is valid unless all n are equal but, more generically, different n and same m can be possible, provided that the intersection of the two most distant local slots is >= requested width. (See figure below, in the case where m = mreq then -> nb = na).
[cyrils' summary of key aspects:
[Editors' note: text from draft-li-ccamp-grid-property-lmp-01]
Potential interworking problems between fixed-grid DWDM and flexible-grid DWDM nodes, may appear. Additionally, even two flexible-grid optical nodes may have different grid properties, leading to link property conflict.
Devices or applications that make use of the flexible-grid may not be able to support every possible slot width. In other words, applications may be defined where different grid granularity can be supported. Taking node F as an example, an application could be defined where the nominal central frequency granularity is 12.5 GHz requiring slot widths being multiple of 25 GHz. Therefore the link between two optical nodes with different grid granularity must be configured to align with the larger of both granularities. Besides, different nodes may have different slot width tuning ranges.
In summary, in a DWDM Link between two nodes, at least the following properties should be negotiated:
Much like in WSON, in which if there is no (available) wavelength converters in an optical network, an LSP is subject to the ''wavelength continuity constraint'' (see section 4 of [RFC6163]), if the capability of shifting or converting an allocated frequency slot, the LSP is subject to the Optical ''Spectrum Continuity Constraint''.
Because of the limited availability of wavelength/spectrum converters (sparse translucent optical network) the wavelength/spectrum continuity constraint should always be considered. When available, information regarding spectrum conversion capabilities at the optical nodes may be used by RSA mechanisms.
The RSA process determines a route and frequency slot for a LSP. Hence, when a route is computed the spectrum assignment process (SA) should determine the central frequency and slot width based on the slot width and available central frequencies information of the transmitter and receiver, and the available frequency ranges information and available slot width ranges of the links that the route traverses.
Similar to RWA for fixed grids, different ways of performing RSA in conjunction with the control plane can be considered. The approaches included in this document are provided for reference purposes only; other possible options could also be deployed.
In this case, a computation entity performs both routing and frequency slot assignment. The computation entity should have the detailed network information, e.g. connectivity topology constructed by nodes/links information, available frequency ranges on each link, node capabilities, etc.
The computation entity could reside either on a PCE or the ingress node.
In this case, routing computation and frequency slot assignment are performed by different entities. The first entity computes the routes and provides them to the second entity; the second entity assigns the frequency slot.
The first entity should get the connectivity topology to compute the proper routes; the second entity should get the available frequency ranges of the links and nodes' capabilities information to assign the spectrum.
In this case, one entity computes the route but the frequency slot assignment is performed hop-by-hop in a distributed way along the route. The available central frequencies which meet the spectrum continuity constraint should be collected hop by hop along the route. This procedure can be implemented by the GMPLS signaling protocol.
In the case of combined RSA architecture, the computation entity needs to get the detailed network information, i.e. connectivity topology, node capabilities and available frequency ranges of the links. Route computation is performed based on the connectivity topology and node capabilities; spectrum assignment is performed based on the available frequency ranges of the links [Editors note: alternatively, it can based the available media channels] . The computation entity may get the detailed network information by the GMPLS routing protocol. Compared with [RFC6163], except wavelength-specific availability information, the connectivity topology and node capabilities are the same as WSON, which can be advertised by GMPLS routing protocol (refer to section 6.2 of [RFC6163]. This section analyses the necessary changes on link information brought by flexible grids.
In the case of flexible grids, channel central frequencies span from 193.1 THz towards both ends of the C band spectrum with 6.25 GHz granularity. Different LSPs could make use of different slot widths on the same link. Hence, the available frequency ranges should be advertised.
The available slot width ranges needs to be advertised, in combination with the Available frequency ranges, in order to verify whether a LSP with a given slot width can be set up or not; this is is constrained by the available slot width ranges of the media matrix Depending on the availability of the slot width ranges, it is possible to allocate more spectrum than strictly needed by the LSP.
[Editors' note: the part on the hierarchy of the optical spectrum could be confusing, we can discuss it]. The total available spectrum on a fiber could be described as a resource that can be divided by a media device into a set of Frequency Slots. In terms of managing spectrum, it is necessary to be able to speak about different granularities of managed spectrum. For example, a part of the spectrum could be assigned to a third party to manage. This need to partition creates the impression that spectrum is a hierarchy in view of Management and Control Plane. The hierarchy is created within a management system, and it is an access right hierarchy only. It is a management hierarchy without any actual resource hierarchy within fiber. The end of fiber is a link end and presents a fiber port which represents all of spectrum available on the fiber. Each spectrum allocation appears as Link Channel Port (i.e., frequency slot port) within fiber.
Fixed DM grids can also be described via suitable choices of slots in a flexible DWDM grid. However, devices or applications that make use of the flexible grid may not be capable of supporting every possible slot width or central frequency position. Following is the definition of information model, not intended to limit any IGP encoding implementation. For example, information required for routing/path selection may be the set of available nominal central frequencies from which a frequency slot of the required width can be allocated. A convenient encoding for this information (may be as a frequency slot or sets of contiguous slices) is further study in IGP encoding document.
<Available Spectrum in Fiber for frequency slot> ::=
<Available Frequency Range-List>
<Available Central Frequency Granularity >
<Available Slot Width Granularity>
<Minimal Slot Width>
<Maximal Slot Width>
<Available Frequency Range-List> ::=
<Available Frequency Range >[< Available Frequency Range-List>]
<Available Frequency Range >::=
<Start Spectrum Position><End Spectrum Position> |
<Sets of contiguous slices>
<Available Central Frequency Granularity> ::= n × 6.25GHz,
where n is positive integer, such as 6.25GHz, 12.5GHz, 25GHz, 50GHz
or 100GHz
<Available Slot Width Granularity> ::= m × 12.5GHz,
where m is positive integer
<Minimal Slot Width> ::= j x 12.5GHz,
j is a positive integer
<Maximal Slot Width> ::= k x 12.5GHz,
k is a positive integer (k >= j)
Figure 18: Routing Information model
[Editor's note: to be discussed]
FFS. Postpone procedures such as resizing existing LSP(s) without deletion, which refers to increase or decrease of slot width value 'm' without changing the value of 'n', etc. until requirements have been identified. At present no hitless resizing protocol has been defined for OCh. Hitless resizing is defined for an ODU entity only.
+=======================================+
| WSON | SSON |
+=======================================+
| OCh | OCh | Signal Layer
+------------------+--------------------+
| | Frequency Slot |
| Optical Channel | | Media Layer
| Carrier | |
+------------------+--------------------+
| 1:1 | N:1 | Relationship
| single layer | MRN/MLN | SL : ML
| network | (* TBD) |
+------------------+--------------------+
Figure 19: Table Comparison WSON/SSON
A SSON (network) refers to the GMPLS control of flexi-grid enabled DWDM optical networks and it encompasses both the signal and media layers. The WSON also encompasses the signal and media layers but, since there is no formal separation between OCh and OCC (1:1) this layer separation is often not considered. A WSON is a particular case of SSON in the which all slot widths are equal and depend on the channel spacing. In other words, since there is only a 1:1 relationship between OCh : OCC there is no need to have separate controlled layers, as if both layers are collapsed into one.
Arguments for LSC switching capability
[QW] A LSP for an optical signal which has a bandwidth of 50GHz passes through both a fixed grid network and a flexible grid network. We assume that no OEOs exist in the LSP, so both the fixed grid path and flexible grid path occupy 50GHz. From the perspective of data plane, there is no change of the signal and no multiplexing when the fixed grid path interconnects with flexible grid path. From this scenario we can conclude that both fixed grid network path and flexible grid network path belong to the same layer. No notion of hierarchy exists between them.
[QW] stitching LSP which is described in [RFC5150] can be applied in one layer. LSP hierarchy allows more than one LSP to be mapped to an H-LSP, but in case of S-LSP, at most one LSP may be associated with an S-LSP. This is similar to the scenario of interconnection between fixed grid LSP and flexible grid LSP. Similar to an H-LSP, an S-LSP could be managed and advertised, although it is not required, as a TE link, either in the same TE domain as it was provisioned or a different one. Path setup procedure of stitching LSP can be applied in the scenario of interconnection between fixed grid path and flexible grid path.
e2e LSP
+++++++++++++++++++++++++++++++++++> (LSP1-2)
LSP segment (flexi-LSP)
====================> (LSP-AB)
C --- E --- G
/|\ | / |\
/ | \ | / | \
R1 ---- A \ | \ | / | / B --- R2
\| \ |/ |/
D --- F --- H
fixed grid --A-- flexi-grid --B-- fixed grid
Figure 20: LSP Stitching [RFC5150] and relationship with fixed-flexi
TBD
The authors would like to thank Pete Anslow for his insights and clarifications.