Network Working Group                                   JL. Le Roux, Ed.
Request for Comments: 5339                                France Telecom
Category: Informational                            D. Papadimitriou, Ed.
                                                          Alcatel-Lucent
                                                          September 2008

Evaluation of Existing GMPLS Protocols
against Multi-Layer and Multi-Region Networks (MLN/MRN)

Status of This Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Abstract

   This document provides an evaluation of Generalized Multiprotocol
   Label Switching (GMPLS) protocols and mechanisms against the
   requirements for Multi-Layer Networks (MLNs) and Multi-Region
   Networks (MRNs).  In addition, this document identifies areas where
   additional protocol extensions or procedures are needed to satisfy
   these requirements, and provides guidelines for potential extensions.

   1. Introduction ....................................................3
      1.1. Conventions Used in This Document ..........................4
   2. MLN/MRN Requirements Overview ...................................4
   3. Analysis ........................................................5
      3.1. Aspects of Multi-Layer Networks ............................5
           3.1.1. Support for Virtual Network Topology
                  Reconfiguration .....................................5
                  3.1.1.1. Control of FA-LSPs Setup/Release ...........5
                  3.1.1.2. Virtual TE Links ...........................6
                  3.1.1.3. Traffic Disruption Minimization
                           during FA Release ..........................8
                  3.1.1.4. Stability ..................................8
           3.1.2. Support for FA-LSP Attribute Inheritance ............9
           3.1.3. FA-LSP Connectivity Verification ....................9
           3.1.4. Scalability .........................................9
           3.1.5. Operations and Management of the MLN/MRN ...........10
                  3.1.5.1. MIB Modules ...............................10
                  3.1.5.2. OAM .......................................11
      3.2. Specific Aspects of Multi-Region Networks .................12
           3.2.1. Support for Multi-Region Signaling .................12
           3.2.2. Advertisement of Adjustment Capacities .............13
   4. Evaluation Conclusion ..........................................16
      4.1. Traceability of Requirements ..............................16
   5. Security Considerations ........................................20
   6. Acknowledgments ................................................20
   7. References .....................................................21
      7.1. Normative References ......................................21
      7.2. Informative References ....................................21
   8. Contributors' Addresses ........................................23

1. Introduction

   Generalized MPLS (GMPLS) extends MPLS to handle multiple switching
   technologies: packet switching, layer-2 switching, TDM (Time Division
   Multiplexing) switching, wavelength switching, and fiber switching
   (see [RFC3945]).  The Interface Switching Capability (ISC) concept is
   introduced for these switching technologies and is designated as
   follows: PSC (Packet Switch Capable), L2SC (Layer-2 Switch Capable),
   TDM capable, LSC (Lambda Switch Capable), and FSC (Fiber Switch
   Capable).  The representation, in a GMPLS control plane, of a
   switching technology domain is referred to as a region [RFC4206].  A
   switching type describes the ability of a node to forward data of a
   particular data plane technology, and uniquely identifies a network
   region.

   A data plane switching layer describes a data plane switching
   granularity level.  For example, LSC, TDM VC-11 and TDM VC-4-64c are
   three different layers.  [RFC5212] defines a Multi-Layer Network
   (MLN) to be a Traffic Engineering (TE) domain comprising multiple
   data plane switching layers either of the same ISC (e.g., TDM) or
   different ISC (e.g., TDM and PSC) and controlled by a single GMPLS
   control plane instance.  [RFC5212] further defines a particular case
   of MLNs.  A Multi-Region Network (MRN) is defined as a TE domain
   supporting at least two different switching types (e.g., PSC and
   TDM), either hosted on the same device or on different ones, and
   under the control of a single GMPLS control plane instance.

   The objectives of this document are to evaluate existing GMPLS
   mechanisms and protocols ([RFC3945], [RFC4202], [RFC3471], [RFC3473])
   against the requirements for MLNs and MRNs, defined in [RFC5212].
   From this evaluation, we identify several areas where additional
   protocol extensions and modifications are required in order to meet
   these requirements, and we provide guidelines for potential
   extensions.

   A summary of MLN/MRN requirements is provided in Section 2.  Then
   Section 3 evaluates whether current GMPLS protocols and mechanisms
   meet each of these requirements.  When the requirements are not met
   by existing protocols, the document identifies whether the required
   mechanisms could rely on GMPLS protocols and procedure extensions, or
   whether it is entirely out of the scope of GMPLS protocols.

   Note that this document specifically addresses GMPLS control plane
   functionality for MLN/MRN in the context of a single administrative
   control plane partition.  Partitions of the control plane where
   separate layers are under distinct administrative control are for
   future study.

This document uses terminologies defined in [RFC3945], [RFC4206], and
[RFC5212].

1.1. 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 [RFC2119].

2. MLN/MRN Requirements Overview

   Section 5 of [RFC5212] lists a set of functional requirements for
   Multi-Layer/Region Networks (MLN/MRN).  These requirements are
   summarized below, and a mapping with sub-sections of [RFC5212] is
   provided.

   Here is the list of requirements that apply to MLN (and thus to MRN):

   - Support for robust Virtual Network Topology (VNT) reconfiguration.
     This implies the following requirements:

        - Optimal control of Forwarding Adjacency Label Switched Path
          (FA-LSP) setup and release (Section 5.8.1 of [RFC5212]);

        - Support for virtual TE links (Section 5.8.2 of [RFC5212]);

        - Minimization of traffic disruption during FA-LSP release
          (Section 5.5 of [RFC5212]);

        - Stability (Section 5.4 of [RFC5212]);

   - Support for FA-LSP attribute inheritance (Section 5.6 of
     [RFC5212]);

   - Support for FA-LSP data plane connectivity verification (Section
     5.9 of [RFC5212]);

   - MLN Scalability (Section 5.3 of [RFC5212]);

   - MLN Operations and Management (OAM) (Section 5.10 of [RFC5212]);

   Here is the list of requirements that apply to MRN only:

   - Support for Multi-Region signaling (Section 5.7 of [RFC5212]);

   - Advertisement of the adjustment capacity (Section 5.2 of
     [RFC5212]);

3. Analysis

3.1. Aspects of Multi-Layer Networks

3.1.1. Support for Virtual Network Topology Reconfiguration

   A set of lower-layer FA-LSPs provides a Virtual Network Topology
   (VNT) to the upper-layer [RFC5212].  By reconfiguring the VNT (FA-LSP
   setup/release) according to traffic demands between source and
   destination node pairs within a layer, network performance factors
   (such as maximum link utilization and residual capacity of the
   network) can be optimized.  Such optimal VNT reconfiguration implies
   several mechanisms that are analyzed in the following sections.

   Note that the VNT approach is just one possible approach to
   performing inter-layer Traffic Engineering.

3.1.1.1. Control of FA-LSPs Setup/Release

   In a Multi-Layer Network, FA-LSPs are created, modified, and released
   periodically according to the change of incoming traffic demands from
   the upper layer.

   This implies a TE mechanism that takes into account the demands
   matrix, the TE topology, and potentially the current VNT, in order to
   compute and setup a new VNT.

   Several functional building blocks are required to support such a TE
   mechanism:

   - Discovery of TE topology and available resources.

   - Collection of upper-layer traffic demands.

   - Policing and scheduling of VNT resources with regard to traffic
     demands and usage (that is, decision to setup/release FA-LSPs).
     The functional component in charge of this function is called a VNT
     Manager (VNTM) [PCE-INTER].

   - VNT Path Computation according to TE topology, potentially taking
     into account the old (existing) VNT in order to minimize changes.
     The functional component in charge of VNT computation may be
     distributed on network elements or may be performed on an external
     element (such as a Path Computation Element (PCE), [RFC4655]).

   - FA-LSP setup/release.

   GMPLS routing protocols provide TE topology discovery.  GMPLS
   signaling protocols allow setting up/releasing FA-LSPs.

   VNTM functions (resources policing/scheduling, decision to
   setup/release FA-LSPs, FA-LSP configuration) are out of the scope of
   GMPLS protocols.  Such functionalities can be achieved directly on
   layer-border Label Switching Routers (LSRs), or through one or more
   external tools.  When an external tool is used, an interface is
   required between the VNTM and the network elements so as to
   setup/release FA-LSPs.  This could use standard management interfaces
   such as [RFC4802].

   The set of traffic demands of the upper layer is required for the VNT
   Manager to take decisions to setup/release FA-LSPs.  Such traffic
   demands include satisfied demands, for which one or more upper-layer
   LSP have been successfully setup, as well as unsatisfied demands and
   future demands, for which no upper layer LSP has been setup yet.  The
   collection of such information is beyond the scope of GMPLS
   protocols.  Note that it may be partially inferred from parameters
   carried in GMPLS signaling or advertised in GMPLS routing.

   Finally, the computation of FA-LSPs that form the VNT can be
   performed directly on layer-border LSRs or on an external element
   (such as a Path Computation Element (PCE), [RFC4655]), and this is
   independent of the location of the VNTM.

   Hence, to summarize, no GMPLS protocol extensions are required to
   control FA-LSP setup/release.

3.1.1.2. Virtual TE Links

   A virtual TE link is a TE link between two upper layer nodes that is
   not actually associated with a fully provisioned FA-LSP in a lower
   layer.  A virtual TE link represents the potentiality to setup an
   FA-LSP in the lower layer to support the TE link that has been
   advertised.  A virtual TE link is advertised as any TE link,
   following the rules in [RFC4206] defined for fully provisioned TE
   links.  In particular, the flooding scope of a virtual TE link is
   within an IGP area, as is the case for any TE link.

   If an upper-layer LSP attempts (through a signaling message) to make
   use of a virtual TE link, the underlying FA-LSP is immediately
   signaled and provisioned (provided there are available resources in
   the lower layer) in the process known as triggered signaling.

   The use of virtual TE links has two main advantages:

   - Flexibility: allows the computation of an LSP path using TE links
     without needing to take into account the actual provisioning status
     of the corresponding FA-LSP in the lower layer;

   - Stability: allows stability of TE links in the upper layer, while
     avoiding wastage of bandwidth in the lower layer, as data plane
     connections are not established until they are actually needed.

   Virtual TE links are setup/deleted/modified dynamically, according to
   the change of the (forecast) traffic demand, operator's policies for
   capacity utilization, and the available resources in the lower layer.

   The support of virtual TE links requires two main building blocks:

   - A TE mechanism for dynamic modification of virtual TE link
     topology;

   - A signaling mechanism for the dynamic setup and deletion of virtual
     TE links.  Setting up a virtual TE link requires a signaling
     mechanism that allows an end-to-end association between virtual TE
     link end points with the purpose of exchanging link identifiers as
     well as some TE parameters.

   The TE mechanism responsible for triggering/policing dynamic
   modification of virtual TE links is out of the scope of GMPLS
   protocols.

   Current GMPLS signaling does not allow setting up and releasing
   virtual TE links.  Hence, GMPLS signaling must be extended to support
   virtual TE links.

   We can distinguish two options for setting up virtual TE links:

   - The Soft FA approach consists of setting up the FA-LSP in the
     control plane without actually activating cross connections in the
     data plane.  On the one hand, this requires state maintenance on
     all transit LSRs (N square issue), but on the other hand, this may
     allow for some admission control.  Indeed, when a Soft FA is
     activated, the resources may no longer be available for use by
     other Soft FAs that have common links.  These Soft FA will be
     dynamically released, and corresponding virtual TE links will be
     deleted.  The Soft FA LSPs may be setup using procedures similar to
     those described in [RFC4872] for setting up secondary LSPs.

   - The remote association approach simply consists of exchanging
     virtual TE link IDs and parameters directly between TE link end
     points.  This does not require state maintenance on transit LSRs,
     but reduces admission control capabilities.  Such an association
     between virtual TE link end points may rely on extensions to the
     Resource Reservation Protocol - Traffic Engineering (RSVP-TE)
     Automatically Switched Optical Network (ASON) call procedure
     [RFC4974].

   Note that the support of virtual TE links does not require any GMPLS
   routing extension.

3.1.1.3. Traffic Disruption Minimization during FA Release

   Before deleting a given FA-LSP, all nested LSPs have to be rerouted
   and removed from the FA-LSP to avoid traffic disruption.  The
   mechanisms required here are similar to those required for graceful
   deletion of a TE link.  A Graceful TE link deletion mechanism allows
   for the deletion of a TE link without disrupting traffic of TE-LSPs
   that were using the TE link.

   Hence, GMPLS routing and/or signaling extensions are required to
   support graceful deletion of TE links.  This may utilize the
   procedures described in [GR-SHUT]: a transit LSR notifies a head-end
   LSR that a TE link along the path of an LSP is going to be torn down,
   and also withdraws the bandwidth on the TE link so that it is not
   used for new LSPs.

3.1.1.4. Stability

   The stability of upper-layer LSP may be impaired if the VNT undergoes
   frequent changes.  In this context, robustness of the VNT is defined
   as the capability to smooth the impact of these changes and avoid
   their subsequent propagation.

   Guaranteeing VNT stability is out of the scope of GMPLS protocols and
   relies entirely on the capability of the TE and VNT management
   algorithms to minimize routing perturbations.  This requires that the
   algorithms take into account the old VNT when computing a new VNT,
   and try to minimize the perturbation.

   Note that a full mesh of lower-layer LSPs may be created between
   every pair of border nodes between the upper and lower layers.  The
   merit of a full mesh of lower-layer LSPs is that it provides
   stability to the upper-layer routing.  That is, the forwarding table
   used in the upper layer is not impacted if the VNT undergoes changes.
   Further, there is always full reachability and immediate access to
   bandwidth to support LSPs in the upper layer.  But it also has

   significant drawbacks, since it requires the maintenance of n^2
   RSVP-TE sessions (where n is the number of border nodes), which may
   be quite CPU- and memory-consuming (scalability impact).  Also, this
   may lead to significant bandwidth wastage.  Note that the use of
   virtual TE links solves the bandwidth wastage issue, and may reduce
   the control plane overload.

3.1.2. Support for FA-LSP Attribute Inheritance

   When an FA TE Link is advertised, its parameters are inherited from
   the parameters of the FA-LSP, and specific inheritance rules are
   applied.

   This relies on local procedures and policies and is out of the scope
   of GMPLS protocols.  Note that this requires that both head-end and
   tail-end of the FA-LSP are driven by same policies.

3.1.3. FA-LSP Connectivity Verification

   Once fully provisioned, FA-LSP liveliness may be achieved by
   verifying its data plane connectivity.

   FA-LSP connectivity verification relies on technology-specific
   mechanisms (e.g., for SDH using G.707 and G.783; for MPLS using
   Bidrectional Forwarding Detection (BFD); etc.) as for any other LSP.
   Hence, this requirement is out of the scope of GMPLS protocols.

   The GMPLS protocols should provide mechanisms for the coordination of
   data link verification in the upper-layer network where data links
   are lower-layer LSPs.

      o GMPLS signaling allows an LSP to be put into 'test' mode
        [RFC3473].
      o The Link Management Protocol [RFC4204] is a targeted protocol
        and can be run end-to-end across lower-layer LSPs.
      o Coordination of testing procedures in different layers is an
        operational matter.

3.1.4. Scalability

   As discussed in [RFC5212]), MRN/MLN routing mechanisms must be
   designed to scale well with an increase of any of the following:
      - Number of nodes
      - Number of TE links (including FA-LSPs)
      - Number of LSPs
      - Number of regions and layers
      - Number of Interface Switching Capability Descriptors (ISCDs) per
        TE link.

   GMPLS routing provides the necessary advertisement functions and is
   based on IETF-designed IGPs.  These are known to scale relatively
   well with the number of nodes and links.  Where there are multiple
   regions or layers, there are two possibilities.

      1.  If a single routing instance distributes information about
         multiple network layers, the effect is no more than to increase
         the number of nodes and links in the network.

      2.  If the MLN is fully integrated (i.e., constructed from hybrid
         nodes), there is an increase in the number of nodes and links
         (as just mentioned), and also a potential increase in the
         amount of ISCD information advertised per link.  This is a
         relatively small amount of information (e.g., 36 bytes in OSPF
         [RFC4203]) per switching type, and each interface is unlikely
         to have more than two or three switching types.

   The number of LSPs in a lower layer that are advertised as TE links
   may impact the scaling of the routing protocol.  A full mesh of FA-
   LSPs in the lower layer would lead to n^2 TE links, where n is the
   number of layer-border LSRs.  This must be taken into consideration
   in the VNT management process.  This is an operational matter beyond
   the scope of GMPLS protocols.

   Since it requires the maintenance of n^2 RSVP-TE sessions (which may
   be quite CPU- and memory-consuming), a full mesh of LSPs in the lower
   layer may impact the scalability of GMPLS signaling.  The use of
   virtual TE links may reduce the control plane overload (see Section
   3.1.1.2).

3.1.5. Operations and Management of the MLN/MRN

   [RFC5212] identifies various requirements for effective management
   and operation of the MLN.  Some features already exist within the
   GMPLS protocol set, some more are under development, and some
   requirements are not currently addressed and will need new
   development work in order to support them.

3.1.5.1. MIB Modules

   MIB modules have been developed to model and control GMPLS switches
   [RFC4803] and to control and report on the operation of the signaling
   protocol [RFC4802].  These may be successfully used to manage the
   operation of a single instance of the control plane protocols that
   operate across multiple layers.

   [RFC4220] provides a MIB module for managing TE links, and this may
   be particularly useful in the context of the MLN because LSPs in the
   lower layers are made available as TE links in the higher layer.

   The traffic engineering database provides a repository for all
   information about the existence and current status of TE links within
   a network.  This information is typically flooded by the routing
   protocol operating within the network, and is used when LSP routes
   are computed.  [TED-MIB] provides a way to inspect the TED to view
   the TE links at the different layers of the MLN.

   As observed in [RFC5212], although it would be possible to manage the
   MLN using only the existing MIB modules, a further MIB module could
   be produced to coordinate the management of separate network layers
   in order to construct a single MLN entity.  Such a MIB module would
   effectively link together entries in the MIB modules already
   referenced.

3.1.5.2. OAM

   At the time of writing, the development of OAM tools for GMPLS
   networks is at an early stage.  GMPLS OAM requirements are addressed
   in [GMPLS-OAM].

   In general, the lower layer network technologies contain their own
   technology-specific OAM processes (for example, SDH/SONET, Ethernet,
   and MPLS).  In these cases, it is not necessary to develop additional
   OAM processes, but GMPLS procedures may be desirable to coordinate
   the operation and configuration of these OAM processes.

   [ETH-OAM] describes some early ideas for this function, but more work
   is required to generalize the technique to be applicable to all
   technologies and to MLN.  In particular, an OAM function operating
   within a server layer must be controllable from the client layer, and
   client layer control plane mechanisms must map and enable OAM in the
   server layer.

   Where a GMPLS-controlled technology does not contain its own OAM
   procedures, this is usually because the technology cannot support
   in-band OAM (for example, Wavelength Division Multiplexing (WDM)
   networks).  In these cases, there is very little that a control plane
   can add to the OAM function since the presence of a control plane
   cannot make any difference to the physical characteristics of the
   data plane.  However, the existing GMPLS protocol suite does provide
   a set of tools that can help to verify the data plane through the
   control plane.  These tools are equally applicable to network
   technologies that do contain their own OAM.

   - Route recording is available through the GMPLS signaling protocol
     [RFC3473], making it possible to check the route reported by the
     control plane against the expected route.  This mechanism also
     includes the ability to record and report the interfaces and labels
     used for the LSP at each hop of its path.

   - The status of TE links is flooded by the GMPLS routing protocols
     [RFC4203] and [RFC4205] making it possible to detect changes in the
     available resources in the network as an LSP is set up.

   - The GMPLS signaling protocol [RFC3473] provides a technique to
     place an LSP into a "test" mode so that end-to-end characteristics
     (such as power levels) may be sampled and modified.

   - The Link Management Protocol [RFC4204] provides a mechanism for
     fault isolation on an LSP.

   - GMPLS signaling [RFC3473] provides a Notify message that can be
     used to report faults and issues across the network.  The message
     includes scaling features to allow one message to report the
     failure of multiple LSPs.

   - Extensions to GMPLS signaling [RFC4783] enable alarm information to
     be collected and distributed along the path of an LSP for more easy
     coordination and correlation.

3.2. Specific Aspects of Multi-Region Networks

3.2.1. Support for Multi-Region Signaling

   There are actually several cases where a transit node could choose
   between multiple Switching Capabilities (SCs) to be used for a
   lower-region FA-LSP:

   - Explicit Route Object (ERO) expansion with loose hops: The transit
     node has to expand the path, and may have to select among a set of
     lower-region SCs.

   - Multi-SC TE link: When the ERO of an FA LSP, included in the ERO of
     an upper-region LSP, comprises a multi-SC TE link, the region
     border node has to select among these SCs.

   Existing GMPLS signaling procedures do not allow solving this
   ambiguous choice of the SC that may be used along a given path.

   Hence, an extension to GMPLS signaling has to be defined to indicate
   the SC(s) that can be used and the SC(s) that cannot be used along
   the path.

3.2.2. Advertisement of Adjustment Capacities

   In the MRN context, nodes supporting more than one switching
   capability on at least one interface are called hybrid nodes
   [RFC5212].  Conceptually, hybrid nodes can be viewed as containing at
   least two distinct switching elements interconnected by internal
   links that provide adjustment between the supported switching
   capabilities.  These internal links have finite capacities and must
   be taken into account when computing the path of a multi-region TE-
   LSP.  The advertisement of the adjustment capacities is required, as
   it provides critical information when performing multi-region path
   computation.

   The term "adjustment capacity" refers to the property of a hybrid
   node to interconnect different switching capabilities it provides
   through its external interfaces [RFC5212].  This information allows
   path computation to select an end-to-end multi-region path that
   includes links of different switching capabilities that are joined by
   LSRs that can adapt the signal between the links.

   Figure 1a below shows an example of a hybrid node.  The hybrid node
   has two switching elements (matrices), which support TDM and PSC
   switching, respectively.  The node has two PSC and TDM ports (Port1
   and Port2, respectively).  It also has an internal link connecting
   the two switching elements.

   The two switching elements are internally interconnected in such a
   way that it is possible to terminate some of the resources of the TDM
   Port2; also, they can provide adjustment of PSC traffic that is
   received/sent over the internal PSC interface (#b).  Two ways are
   possible to set up PSC LSPs (Port1 or Port2).  Available resources
   advertisement (e.g., Unreserved and Min/Max LSP Bandwidth) should
   cover both ways.

                             Network element
                        .............................
                        :            --------       :
              PSC       :           |  PSC   |      :
            Port1-------------<->---|#a      |      :
                        :  +--<->---|#b      |      :
                        :  |         --------       :
                        :  |        ----------      :
              TDM       :  +--<->--|#c  TDM   |     :
            Port2 ------------<->--|#d        |     :
                        :           ----------      :
                        :............................

                           Figure 1a.  Hybrid node.

   Port1 and Port2 can be grouped together thanks to internal Dense
   Wavelength Division Multiplexing (DWDM), to result in a single
   interface: Link1.  This is illustrated in Figure 1b below.

                             Network element
                        .............................
                        :            --------       :
                        :           |  PSC   |      :
                        :           |        |      :
                        :         --|#a      |      :
                        :        |  |   #b   |      :
                        :        |   --------       :
                        :        |       |          :
                        :        |  ----------      :
                        :    /|  | |    #c    |     :
                        :   | |--  |          |     :
              Link1 ========| |    |    TDM   |     :
                        :   | |----|#d        |     :
                        :    \|     ----------      :
                        :............................

                           Figure 1b.  Hybrid node.

   Let's assume that all interfaces are STM16 (with VC4-16c capable as
   Max LSP bandwidth).  After setting up several PSC LSPs via port #a
   and setting up and terminating several TDM LSPs via port #d and port
   #b, a capacity of only 155 Mb is still available on port #b.
   However, a 622 Mb capacity remains on port #a, and VC4-5c capacity
   remains on port #d.

   When computing the path for a new VC4-4c TDM LSP, one must know that
   this node cannot terminate this LSP, as there is only a 155 Mb
   capacity still available for TDM-PSC adjustment.  Hence, the TDM-PSC
   adjustment capacity must be advertised.

   With current GMPLS routing [RFC4202], this advertisement is possible
   if link bundling is not used and if two TE links are advertised for
   Link1.

   We would have the following TE link advertisements:

   TE link 1 (Port1):
     - ISCD sub-TLV: PSC with Max LSP bandwidth = 622 Mb
     - Unreserved bandwidth = 622 Mb.

   TE link 2 (Port2):
     - ISCD #1 sub-TLV: TDM with Max LSP bandwidth = VC4-4c,
     - ISCD #2 sub-TLV: PSC with Max LSP bandwidth = 155 Mb,
     - Unreserved bandwidth (equivalent): 777 Mb.

   The ISCD #2 in TE link 2 actually represents the TDM-PSC adjustment
   capacity.

   However, if for obvious scalability reasons, link bundling is done,
   then the adjustment capacity information is lost with current GMPLS
   routing, as we have the following TE link advertisement:

   TE link 1 (Port1 + Port2):
     - ISCD #1 sub-TLV: TDM with Max LSP bandwidth = VC4-4c,
     - ISCD #2 sub-TLV: PSC with Max LSP bandwidth = 622 Mb,
     - Unreserved bandwidth (equivalent): 1399 Mb.

   With such a TE link advertisement, an element computing the path of a
   VC4-4c LSP cannot know that this LSP cannot be terminated on the
   node.

   Thus, current GMPLS routing can support the advertisement of the
   adjustment capacities, but this precludes performing link bundling
   and thus faces significant scalability limitations.

   Hence, GMPLS routing must be extended to meet this requirement.  This
   could rely on the advertisement of the adjustment capacities as a new
   TE link attribute (that would complement the Interface Switching
   Capability Descriptor TE link attribute).

   Note: Multiple ISCDs MAY be associated with a single switching
   capability.  This can be performed to provide (e.g., for TDM
   interfaces) the Min/Max LSP Bandwidth associated to each layer (or
   set of layers) for that switching capability.  For example, an
   interface associated to TDM switching capability and supporting VC-12
   and VC-4 switching can be associated to one ISCD sub-TLV or two ISCD
   sub-TLVs.  In the first case, the Min LSP Bandwidth is set to VC-12
   and the Max LSP Bandwidth to VC-4.  In the second case, the Min LSP
   Bandwidth is set to VC-12 and the Max LSP Bandwidth to VC-12, in the
   first ISCD sub-TLV; and the Min LSP Bandwidth is set to VC-4 and the
   Max LSP Bandwidth to VC-4, in the second ISCD sub-TLV.  Hence, in the
   first case, as long as the Min LSP Bandwidth is set to VC-12 (and not
   VC-4), and in the second case, as long as the first ISCD sub-TLV is
   advertised, there is sufficient capacity across that interface to
   setup a VC-12 LSP.

4. Evaluation Conclusion

   Most of the required MLN/MRN functions will rely on mechanisms and
   procedures that are out of the scope of the GMPLS protocols, and thus
   do not require any GMPLS protocol extensions.  They will rely on
   local procedures and policies, and on specific TE mechanisms and
   algorithms.

   As regards Virtual Network Topology (VNT) computation and
   reconfiguration, specific TE mechanisms need to be defined, but these
   mechanisms are out of the scope of GMPLS protocols.

   Six areas for extensions of GMPLS protocols and procedures have been
   identified:

   - GMPLS signaling extension for the setup/deletion of the virtual TE
     links;

   - GMPLS signaling extension for graceful TE link deletion;

   - GMPLS signaling extension for constrained multi-region signaling
     (SC inclusion/exclusion);

   - GMPLS routing extension for the advertisement of the adjustment
     capacities of hybrid nodes.

   - A MIB module for coordination of other MIB modules being operated
     in separate layers.

   - GMPLS signaling extensions for the control and configuration of
     technology-specific OAM processes.

4.1. Traceability of Requirements

   This section provides a brief cross-reference to the requirements set
   out in [RFC5212] so that it is possible to verify that all of the
   requirements listed in that document have been examined in this
   document.

   - Path computation mechanism should be able to compute paths and
     handle topologies consisting of any combination of (simplex) nodes
     ([RFC5212], Section 5.1).
     o Path computation mechanisms are beyond the scope of protocol
     specifications, and out of scope for this document.

   - A hybrid node should maintain resources on its internal links
     ([RFC5212], Section 5.2).
     o This is an implementation requirement and is beyond the scope of
     protocol specifications, and it is out of scope for this document.

   - Path computation mechanisms should be prepared to use the
     availability of termination/adjustment resources as a constraint in
     path computation ([RFC5212], Section 5.2).
     o Path computation mechanisms are beyond the scope of protocol
     specifications, and out of scope for this document.

   - The advertisement of a node's ability to terminate lower-region
     LSPs and to forward traffic in the upper-region (adjustment
     capability) is required ([RFC5212], Section 5.2).
     o See 3.2.2">Section 3.2.2 of this document.

   - The path computation mechanism should support the coexistence of
     upper-layer links directly connected to upper-layer switching
     elements, and upper-layer links connected through internal links
     between upper-layer and lower-layer switching elements ([RFC5212],
     Section 5.2).
     o Path computation mechanisms are beyond the scope of protocol
     specifications, and out of scope for this document.

   - MRN/MLN routing mechanisms must be designed to scale well with an
     increase of any of the following:
     - Number of nodes
     - Number of TE links (including FA-LSPs)
     - Number of LSPs
     - Number of regions and layers
     - Number of ISCDs per TE link.
     ([RFC5212], Section 5.3).
     o See 3.1.4">Section 3.1.4 of this document.

   - Design of the routing protocols must not prevent TE information
     filtering based on ISCDs ([RFC5212], Section 5.3).
     o All advertised information carries the ISCD, and so a receiving
     node may filter as required.

   - The path computation mechanism and the signaling protocol should be
     able to operate on partial TE information, ([RFC5212], Section
     5.3).
     o Path computation mechanisms are beyond the scope of protocol
     specifications, and out of scope for this document.

   - Protocol mechanisms must be provided to enable creation, deletion,
     and modification of LSPs triggered through operational actions
     ([RFC5212], Section 5.4).
     o Such mechanisms are standard in GMPLS signaling [RFC3473].

   - Protocol mechanisms should be provided to enable similar functions
     triggered by adjacent layers ([RFC5212], Section 5.4).
     o Such mechanisms are standard in GMPLS signaling [RFC3473].

   - Protocol mechanisms may be provided to enable adaptation to changes
     such as traffic demand, topology, and network failures.  Routing
     robustness should be traded with adaptability of those changes
     ([RFC5212], Section 5.4).
     o See 3.1.1">Section 3.1.1 of this document.

   - Reconfiguration of the VNT must be as non-disruptive as possible
     and must be under the control of policy configured by the operator
     ([RFC5212], Section 5.5).
     o See 3.1.1.3">Section 3.1.1.3 of this document

   - Parameters of a TE link in an upper layer should be inherited from
     the parameters of the lower-layer LSP that provides the TE link,
     based on polices configured by the operator ([RFC5212], Section
     5.6).
     o See 3.1.2">Section 3.1.2 of this document.

   - The upper-layer signaling request may contain an ERO that includes
     only hops in the upper layer ([RFC5212], Section 5.7).
     o Standard for GMPLS signaling [RFC3473].  See also Section 3.2.1.

   - The upper-layer signaling request may contain an ERO specifying the
     lower layer FA-LSP route ([RFC5212], Section 5.7).
     o Standard for GMPLS signaling [RFC3473].  See also Section 3.2.1.

   - As part of the re-optimization of the MLN, it must be possible to
     reroute a lower-layer FA-LSP while keeping interface identifiers of
     the corresponding TE links unchanged and causing only minimal
     disruption to higher-layer traffic ([RFC5212], Section 5.8.1).
     o See 3.1.1.3">Section 3.1.1.3.

   - The solution must include measures to protect against network
     destabilization caused by the rapid setup and tear-down of lower-
     layer LSPs, as traffic demand varies near a threshold ([RFC5212],
     Sections 5.8.1 and 5.8.2).
     o See Section 3.1.1.4.

   - Signaling of lower-layer LSPs should include a mechanism to rapidly
     advertise the LSP as a TE link in the upper layer, and to
     coordinate into which routing instances the TE link should be
     advertised ([RFC5212], Section 5.8.1).
     o This is provided by [RFC4206] and enhanced by [HIER-BIS].  See
     also Section 3.1.1.2.

   - If an upper-layer LSP is set up making use of a virtual TE link,
     the underlying LSP must immediately be signaled in the lower layer
     ([RFC5212], Section 5.8.2).
     o See 3.1.1.2">Section 3.1.1.2.

   - The solution should provide operations to facilitate the build-up
     of virtual TE links, taking into account the forecast upper-layer
     traffic demand, and available resource in the lower layer
     ([RFC5212], Section 5.8.2).
     o See 3.1.1.2">Section 3.1.1.2 of this document.

   - The GMPLS protocols should provide mechanisms for the coordination
     of data link verification in the upper-layer network where data
     links are lower layer LSPs ([RFC5212], Section 5.9).
     o See 3.1.3">Section 3.1.3 of this document.

   - Multi-layer protocol solutions should be manageable through MIB
     modules ([RFC5212], Section 5.10).
     o See 3.1.5.1">Section 3.1.5.1.

   - Choices about how to coordinate errors and alarms, and how to
     operate OAM across administrative and layer boundaries must be left
     open for the operator ([RFC5212], Section 5.10).
     o This is an implementation matter, subject to operational
     policies.

   - It must be possible to enable end-to-end OAM on an upper-layer LSP.
     This function appears to the ingress LSP as normal LSP-based OAM
     [GMPLS-OAM], but at layer boundaries, depending on the technique
     used to span the lower layers, client-layer OAM operations may need
     to be mapped to server-layer OAM operations ([RFC5212], Section
     5.10).
     o See 3.1.5.2">Section 3.1.5.2.

   - Client-layer control plane mechanisms must map and enable OAM in
     the server layer ([RFC5212], Section 5.10).
     o See 3.1.5.2">Section 3.1.5.2.

   - OAM operation enabled for an LSP in a client layer must operate for
     that LSP along its entire length ([RFC5212], Section 5.10).
     o See 3.1.5.2">Section 3.1.5.2.

   - OAM function operating within a server layer must be controllable
     from the client layer.  Such control should be subject to policy at
     the layer boundary ([RFC5212], Section 5.10).
     o This is an implementation matter.

   - The status of a server layer LSP must be available to the client
     layer.  This information should be configurable to be automatically
     notified to the client layer at the layer boundary, and should be
     subject to policy ([RFC5212], Section 5.10).
     o This is an implementation matter.

   - Implementations may use standardized techniques (such as MIB
     modules) to convey status information between layers.
     o This is an implementation matter.

5. Security Considerations

   [RFC5212] sets out the security requirements for operating a MLN or
   MRN.  These requirements are, in general, no different from the
   security requirements for operating any GMPLS network.  As such, the
   GMPLS protocols already provide adequate security features.  An
   evaluation of the security features for GMPLS networks may be found
   in [MPLS-SEC], and where issues or further work is identified by that
   document, new security features or procedures for the GMPLS protocols
   will need to be developed.

   [RFC5212] also identifies that where the separate layers of a MLN/MRN
   are operated as different administrative domains, additional security
   considerations may be given to the mechanisms for allowing inter-
   layer LSP setup.  However, this document is explicitly limited to the
   case where all layers under GMPLS control are part of the same
   administrative domain.

   Lastly, as noted in [RFC5212], it is expected that solution documents
   will include a full analysis of the security issues that any protocol
   extensions introduce.

6. Acknowledgments

   We would like to thank Julien Meuric, Igor Bryskin, and Adrian Farrel
   for their useful comments.

   Thanks also to Question 14 of Study Group 15 of the ITU-T for their
   thoughtful review.

7. References

7.1. Normative References

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

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

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

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

   [RFC5212]   Shiomoto, K., Papadimitriou, D., Le Roux, JL., Vigoureux,
               M., and D. Brungard, "Requirements for GMPLS-Based
               Multi-Region and Multi-Layer Networks (MRN/MLN)", RFC
               5212, July 2008.

7.2. Informative References

   [RFC3473]   Berger, L., Ed., "Generalized Multi-Protocol Label
               Switching (GMPLS) Signaling Resource ReserVation
               Protocol-Traffic Engineering (RSVP-TE) Extensions", RFC
               3473, January 2003.

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

   [RFC4204]   Lang, J., Ed., "Link Management Protocol (LMP)", RFC
               4204, October 2005.

   [RFC4205]   Kompella, K., Ed., and Y. Rekhter, Ed., "Intermediate
               System to Intermediate System (IS-IS) Extensions in
               Support of Generalized Multi-Protocol Label Switching
               (GMPLS)", RFC 4205, October 2005.

   [RFC4206]   Kompella, K. and Y. Rekhter, "Label Switched Paths (LSP)
               Hierarchy with Generalized Multi-Protocol Label Switching
               (GMPLS) Traffic Engineering (TE)", RFC 4206, October
               2005.

   [RFC4220]   Dubuc, M., Nadeau, T., and J. Lang, "Traffic Engineering
               Link Management Information Base", RFC 4220, November
               2005.

   [RFC4655]   Farrel, A., Vasseur, J.-P., and J. Ash, "A Path
               Computation Element (PCE)-Based Architecture", RFC 4655,
               August 2006.

   [RFC4783]   Berger, L., Ed., "GMPLS - Communication of Alarm
               Information", RFC 4783, December 2006.

   [RFC4802]   Nadeau, T., Ed., and A. Farrel, Ed., "Generalized
               Multiprotocol Label Switching (GMPLS) Traffic Engineering
               Management Information Base", RFC 4802, February 2007.

   [RFC4803]   Nadeau, T., Ed., and A. Farrel, Ed., "Generalized
               Multiprotocol Label Switching (GMPLS) Label Switching
               Router (LSR) Management Information Base", RFC 4803,
               February 2007.

   [RFC4872]   Lang, J., Ed., Rekhter, Y., Ed., and D. Papadimitriou,
               Ed., "RSVP-TE Extensions in Support of End-to-End
               Generalized Multi-Protocol Label Switching (GMPLS)
               Recovery", RFC 4872, May 2007.

   [RFC4974]   Papadimitriou, D. and A. Farrel, "Generalized MPLS
               (GMPLS) RSVP-TE Signaling Extensions in Support of
               Calls", RFC 4974, August 2007.

   [ETH-OAM]   Takacs, A., Gero, B., and D. Mohan, "GMPLS RSVP-TE
               Extensions to Control Ethernet OAM", Work in Progress,
               July 2008.

   [GMPLS-OAM] Nadeau, T., Otani, T. Brungard, D., and A. Farrel, "OAM
               Requirements for Generalized Multi-Protocol Label
               Switching (GMPLS) Networks", Work in Progress, October
               2007.

   [GR-SHUT]   Ali, Z., Zamfir, A., and J. Newton, "Graceful Shutdown in
               MPLS and Generalized MPLS Traffic Engineering Networks",
               Work in Progress, July 2008.

   [HIER-BIS]  Shiomoto, K., Rabbat, R., Ayyangar, A., Farrel, A., and
               Z. Ali, "Procedures for Dynamically Signaled Hierarchical
               Label Switched Paths", Work in Progress, February 2008.

   [MPLS-SEC]  Fang, L., Ed., "Security Framework for MPLS and GMPLS
               Networks", Work in Progress, July 2008.

   [PCE-INTER] Oki, E., Le Roux , J-L., and A. Farrel, "Framework for
               PCE-Based Inter-Layer MPLS and GMPLS Traffic
               Engineering", Work in Progress, June 2008.

   [TED-MIB]   Miyazawa, M., Otani, T., Nadeau, T., and K. Kunaki,
               "Traffic Engineering Database Management Information Base
               in support of MPLS-TE/GMPLS", Work in Progress, July
               2008.

8. Contributors' Addresses

   Deborah Brungard
   AT&T
   Rm. D1-3C22 - 200 S. Laurel Ave.
   Middletown, NJ, 07748 USA
   EMail: dbrungard@att.com

   Eiji Oki
   NTT
   3-9-11 Midori-Cho
   Musashino, Tokyo 180-8585, Japan
   EMail: oki.eiji@lab.ntt.co.jp

   Kohei Shiomoto
   NTT
   3-9-11 Midori-Cho
   Musashino, Tokyo 180-8585, Japan
   EMail: shiomoto.kohei@lab.ntt.co.jp

   M. Vigoureux
   Alcatel-Lucent France
   Route de Villejust
   91620 Nozay
   FRANCE
   EMail: martin.vigoureux@alcatel-lucent.fr

Editors' Addresses

   Jean-Louis Le Roux
   France Telecom
   2, avenue Pierre-Marzin
   22307 Lannion Cedex, France
   EMail: jeanlouis.leroux@orange-ftgroup.com

   Dimitri Papadimitriou
   Alcatel-Lucent
   Francis Wellensplein 1,
   B-2018 Antwerpen, Belgium
   EMail: dimitri.papadimitriou@alcatel-lucent.be

Full 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.

   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.

Intellectual Property

   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.