Network Working Group                                         G. Swallow
Request for Comments: 4208                            Cisco Systems, Inc
Category: Standards Track                                       J. Drake
                                                                  Boeing
                                                            H. Ishimatsu
                                                           G1M Co., Ltd.
                                                              Y. Rekhter
                                                   Juniper Networks, Inc
                                                            October 2005

           Generalized Multiprotocol Label Switching (GMPLS)
                     User-Network Interface (UNI):
      Resource ReserVation Protocol-Traffic Engineering (RSVP-TE)
                     Support for the Overlay Model

Status of This Memo

   This document specifies an Internet standards track protocol for the
   Internet community, and requests discussion and suggestions for
   improvements.  Please refer to the current edition of the "Internet
   Official Protocol Standards" (STD 1) for the standardization state
   and status of this protocol.  Distribution of this memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2005).

Abstract

   Generalized Multiprotocol Label Switching (GMPLS) defines both
   routing and signaling protocols for the creation of Label Switched
   Paths (LSPs) in various switching technologies.  These protocols can
   be used to support a number of deployment scenarios.  This memo
   addresses the application of GMPLS to the overlay model.

Swallow, et al.             Standards Track                     [Page 1]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

Table of Contents

   1. Introduction ....................................................2
      1.1. GMPLS User-Network Interface (GMPLS UNI) ...................4
   2. Addressing ......................................................5
   3. ERO Processing ..................................................6
      3.1. Path Message without ERO ...................................6
      3.2. Path Message with ERO ......................................6
      3.3. Explicit Label Control .....................................7
   4. RRO Processing ..................................................7
   5. Notification ....................................................7
   6. Connection Deletion .............................................8
      6.1. Alarm-Free Connection Deletion .............................8
      6.2. Connection Deletion with PathErr ...........................8
   7. VPN Connections .................................................9
   8. Security Considerations ........................................10
   9. Acknowledgments ................................................10
   10. References ....................................................10
      10.1. Normative References .....................................10
      10.2. Informational References .................................10

1.  Introduction

   Generalized Multiprotocol Label Switching (GMPLS) defines both
   routing and signaling protocols for the creation of Label Switched
   Paths (LSPs) in various transport technologies.  These protocols can
   be used to support a number of deployment scenarios.  In a peer
   model, edge-nodes support both a routing and a signaling protocol.
   The protocol interactions between an edge-node and a core-node are
   the same as between two core-nodes.  In the overlay model, the core-
   nodes act more as a closed system.  The edge-nodes do not participate
   in the routing protocol instance that runs among the core nodes; in
   particular, the edge-nodes are unaware of the topology of the core-
   nodes.  There may, however, be a routing protocol interaction between
   a core-node and an edge-node for the exchange of reachability
   information to other edge-nodes.

Swallow, et al.             Standards Track                     [Page 2]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

     Overlay                                                  Overlay
     Network       +----------------------------------+       Network
   +---------+     |                                  |     +---------+
   |  +----+ |     |  +-----+    +-----+    +-----+   |     | +----+  |
   |  |    | |     |  |     |    |     |    |     |   |     | |    |  |
   | -+ EN +-+-----+--+ CN  +----+ CN  +----+  CN +---+-----+-+ EN +- |
   |  |    | |  +--+--|     |    |     |    |     |   |     | |    |  |
   |  +----+ |  |  |  +--+--+    +--+--+    +--+--+   |     | +----+  |
   |         |  |  |     |          |          |      |     |         |
   +---------+  |  |     |          |          |      |     +---------+
                |  |     |          |          |      |
   +---------+  |  |     |          |          |      |     +---------+
   |         |  |  |  +--+--+       |       +--+--+   |     |         |
   |  +----+ |  |  |  |     |       +-------+     |   |     | +----+  |
   |  |    +-+--+  |  | CN  +---------------+ CN  |   |     | |    |  |
   | -+ EN +-+-----+--+     |               |     +---+-----+-+ EN +- |
   |  |    | |     |  +-----+               +-----+   |     | |    |  |
   |  +----+ |     |                                  |     | +----+  |
   |         |     +----------------------------------+     |         |
   +---------+                Core Network                  +---------+
     Overlay                                                  Overlay
     Network                                                  Network

                        Legend:   EN  -  Edge Node
                                  CN  -  Core Node

                    Figure 1:  Overlay Reference Model

   Figure 1 shows a reference network.  The core network is represented
   by the large box in the center.  It contains five core-nodes marked
   'CN'.  The four boxes around the edge marked "Overlay Network"
   represent four islands of a single overlay network.  Only the nodes
   of this network with TE links into the core network are shown.  These
   nodes are called edge-nodes; the terminology is in respect to the
   core network, not the overlay network.  Note that each box marked
   "Overlay Network" could contain many other nodes.  Such nodes are not
   shown; they do not participate directly in the signaling described in
   this document.  Only the edge-nodes can signal to set up links across
   the core to other edge-nodes.

   How a link between edge-nodes is requested and triggered is out of
   the scope of this document, as is precisely how that link is used by
   the Overlay Network.  One possibility is that the edge-nodes will
   inform the other nodes of the overlay network of the existence of the
   link, possibly using a forwarding adjacency as described in
   [MPLS-HIER].  Note that this contrasts with a forwarding adjacency
   that is provided by the core network as a link between core-nodes.

Swallow, et al.             Standards Track                     [Page 3]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

   In the overlay model, there may be restrictions on what may be
   signaled between an edge-node and a core-node.  This memo addresses
   the application of GMPLS to the overlay model.  Specifically, it
   addresses RSVP-TE procedures between an edge-node and a core-node in
   the overlay model.  All signaling procedures are identical to the
   GMPLS extensions specified in [RFC3473], except as noted in this
   document.

   This document primarily addresses interactions between an edge-node
   and it's adjacent (at the data plane) core-node; out-of-band and
   non-adjacent signaling capabilities may mean that signaling messages
   are delivered on a longer path.  Except where noted, the term core-
   node refers to the node immediately adjacent to an edge-node across a
   particular data plane interface.  The term core-nodes, however,
   refers to all nodes in the core.

   Realization of a single or multiple instance of the UNI is
   implementation dependent at both the CN and EN so long as it meets
   the functional requirements for robustness, security, and privacy
   detailed in Section 7.

   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].

   Readers are assumed to be familiar with the terminology introduced in
   [RFC3031], [GMPLS-ARCH], and [RFC3471].

1.1.  GMPLS User-Network Interface (GMPLS UNI)

   One can apply the GMPLS Overlay model at the User-Network Interface
   (UNI) reference point defined in the Automatically Switched Optical
   Network (ASON) [G.8080].  Consider the case where the 'Core Network'
   in Figure 1 is a Service Provider network, and the Edge Nodes are
   'user' devices.  The interface between an EN and a CN is the UNI
   reference point, and to support the ASON model, one must define
   signaling across the UNI.

   The extensions described in this memo provide mechanisms for UNI
   signaling that are compatible with GMPLS signaling [RFC3471,
   RFC3473].  Moreover, these mechanisms for UNI signaling are in line
   with the RSVP model; namely, there is a single end-to-end RSVP
   session for the user connection.  The first and last hops constitute
   the UNI, and the RSVP session carries the user parameters end-to-end.
   This obviates the need to map (or carry) user parameters to (in) the
   format expected by the network-to-network interface (NNI) used within
   the Service Provider network.  This in turn means that the UNI and
   NNI can be independent of one another, which is a requirement of the

Swallow, et al.             Standards Track                     [Page 4]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

   ASON architecture.  However, in the case that the UNI and NNI are
   both GMPLS RSVP-based, the methodology specified in this memo allows
   for a single RSVP session to instantiate both UNI and NNI signaling,
   if so desired, and if allowed by Service Provider policy.

2.  Addressing

   Addresses for edge-nodes in the overlay model are drawn from the same
   address space as the edge-nodes use to address their adjacent core-
   nodes.  This may be the same address space as used by the core-nodes
   to communicate among themselves, or it may be a VPN space supported
   by the core-nodes as an overlay.

   To be more specific, an edge-node and its attached core-node must
   share the same address space that is used by GMPLS to signal between
   the edge-nodes across the core network.  A set of  tuples share the same address space if the edge-nodes in the
   set could establish LSPs (through the core-nodes) among themselves
   without address mapping or translation (note that edge-nodes in the
   set may be a subset of all the edge-nodes).  The address space used
   by the core-nodes to communicate among themselves may, but need not,
   be shared with the address space used by any of the  tuples.  This does not imply a mandatory 1:1 mapping
   between a set of LSPs and a given addressing space.

   When multiple overlay networks are supported by a single core
   network, one or more address spaces may be used according to privacy
   requirements.  This may be achieved without varying the core-node
   addresses since it is the   tuple that
   constitutes address space membership.

   An edge-node is identified by either a single IP address representing
   its Node-ID, or by one or more numbered TE links that connect the
   edge-node to the core-nodes.  Core-nodes are assumed to be ignorant
   of any other addresses associated with an edge-node (i.e., addresses
   that are not used in signaling connections through the GMPLS core).

   An edge-node need only know its own address, an address of the
   adjacent core-node, and know (or be able to resolve) the address of
   any other edge-node to which it wishes to connect, as well as (of
   course) the addresses used in the overlay network island of which it
   is a part.

   A core-node need only know (and track) the addresses on interfaces
   between that core-node and its attached edge-nodes, as well as the
   Node IDs of those edge-nodes.  In addition, a core-node needs to know
   the interface addresses and Node IDs of other edge-nodes to which an
   attached edge-node is permitted to connect.

Swallow, et al.             Standards Track                     [Page 5]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

   When forming a SENDER_TEMPLATE, the ingress edge-node includes either
   its Node-ID or the address of one of its numbered TE links.  In the
   latter case the connection will only be made over this interface.

   When forming a SESSION_OBJECT, the ingress edge-node includes either
   the Node-ID of the egress edge-device or the address of one of the
   egress' numbered TE links.  In the latter case the connection will
   only be made over this interface.  The Extended_Tunnel_ID of the
   SESSION Object is set to either zero or to an address of the ingress
   edge-device.

   Links may be either numbered or unnumbered.  Further, links may be
   bundled or unbundled.  See [GMPLS-ARCH], [RFC3471], [BUNDLE], and
   [RFC3477].

3. ERO Processing

   An edge-node MAY include an ERO.  A core-node MAY reject a Path
   message that contains an ERO.  Such behavior is controlled by
   (hopefully consistent) configuration.  If a core-node rejects a Path
   message due to the presence of an ERO, it SHOULD return a PathErr
   message with an error code of "Unknown object class" toward the
   sender as described in [RFC3209].  This causes the path setup to
   fail.

   Further, a core-node MAY accept EROs that only include the ingress
   edge-node, the ingress core-node, the egress core-node, and the
   egress edge-node.  This is to support explicit label control on the
   edge-node interface; see below.  If a core-node rejects a Path
   message due to the presence of an ERO that is not of the permitted
   format, it SHOULD return a PathErr message with an error code of Bad
   Explicit Route Object as defined in [RFC3209].

3.1. Path Message without ERO

   When a core-node receives a Path message from an edge-node that
   contains no ERO, it MUST calculate a route to the destination and
   include that route in an ERO, before forwarding the PATH message.
   One exception would be if the egress edge-node were also adjacent to
   this core-node.  If no route can be found, the core-node SHOULD
   return a PathErr message with an error code and value of 24,5 - "No
   route available toward destination".

3.2. Path Message with ERO

   When a core-node receives a Path message from an edge-node that
   contains an ERO, it SHOULD verify the route against its topology
   database before forwarding the PATH message.  If the route is not

Swallow, et al.             Standards Track                     [Page 6]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

   viable (according to topology, currently available resources, or
   local policy), then a PathErr message with an error code and value of
   24,5 - "No route available toward destination" should be returned.

3.3. Explicit Label Control

   In order to support explicit label control and full identification of
   the egress link, an ingress edge-node may include this information in
   the ERO that it passes to its neighboring core-node.  In the case
   that no other ERO is supplied, this explicit control information is
   provided as the only hop of the ERO and is encoded by setting the
   first subobject of the ERO to the node-ID of the egress core-node
   with the L-bit set; following this subobject are all other subobjects
   necessary to identify the link and labels as they would normally
   appear.

   The same rules apply to the presence of the explicit control
   subobjects as the last hop in the ERO, if a fuller ERO is supplied by
   the ingress edge-node to its neighbor core-node; but in this case the
   L-bit MAY be clear.

   This process is described in [RFC3473] and [EXPLICIT].

4. RRO Processing

   An edge-node MAY include an RRO.  A core-node MAY remove the RRO from
   the Path message before forwarding it.  Further, the core-node may
   remove the RRO from a Resv message before forwarding it to the edge-
   node.  Such behavior is controlled by (hopefully consistent)
   configuration.

   Further, a core-node MAY edit the RRO in a Resv message such that it
   includes only the subobjects from the egress core-node through the
   egress edge-node.  This is to allow the ingress node to be aware of
   the selected link and labels on at the far end of the connection.

5. Notification

   An edge-node MAY include a NOTIFY_REQUEST object in both the Path and
   Resv messages it generates.  Core-nodes may send Notify messages to
   edge-nodes that have included the NOTIFY_REQUEST object.

   A core-node MAY remove a NOTIFY_REQUEST object from a Path or Resv
   message received from an edge-node before forwarding it.

   If no NOTIFY_REQUEST object is present in the Path or Resv received
   from an edge-node, the core-node adjacent to the edge-node may
   include a NOTIFY_REQUEST object and set its value to its own address.

Swallow, et al.             Standards Track                     [Page 7]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

   In either of the above cases, the core-node SHOULD NOT send Notify
   messages to the edge-node.

   When a core-node receives a NOTIFY_REQUEST object from an edge-node,
   it MAY update the Notify Node Address with its own address before
   forwarding it.  In this case, when Notify messages are received, they
   MAY be selectively (based on local policy) forwarded to the edge-
   node.

6. Connection Deletion

6.1. Alarm-Free Connection Deletion

   RSVP-TE currently deletes connections using either a single pass
   PathTear message, or a ResvTear and PathTear message combination.
   Upon receipt of the PathTear message, a node deletes the connection
   state and forwards the message.  In optical networks, however, it is
   possible that the deletion of a connection (e.g., removal of the
   cross-connect) in a node may cause the connection to be perceived as
   failed in downstream nodes (e.g., loss of frame, loss of light,
   etc.). This may in turn lead to management alarms and perhaps the
   triggering of restoration/protection for the connection.

   To address this issue, the graceful connection deletion procedure
   SHOULD be followed.  Under this procedure, an ADMIN_STATUS object
   MUST be sent in a Path or Resv message along the connection's path to
   inform all nodes en route to the intended deletion, prior to the
   actual deletion of the connection.  The procedure is described in
   [RFC3473].

   If an ingress core-node receives a PathTear without having first seen
   an ADMIN_STATUS object informing it that the connection is about to
   be deleted, it MAY pause the PathTear and first send a Path message
   with an ADMIN_STATUS object to inform all downstream LSRs that the
   connection is about to be deleted.  When the Resv is received echoing
   the ADMIN_STATUS or using a timer as described in [RFC3473], the
   ingress core-node MUST forward the PathTear.

6.2. Connection Deletion with PathErr

   [RFC3473] introduces the Path_State_Removed flag to a PathErr message
   to indicate that the sender has removed all state associated with the
   LSP and does not need to see a PathTear.  A core-node next to an
   edge-node MAY map between teardown using ResvTear/PathTear and
   PathErr with Path_state_Removed.

Swallow, et al.             Standards Track                     [Page 8]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

   A core-node next to an edge-node receiving a ResvTear from its
   downstream neighbor MAY respond with a PathTear and send a PathErr
   with Path_State_Removed further upstream.

   Note, however, that a core-node next to an edge-node receiving a
   PathErr with Path_State_Removed from its downstream neighbor MUST NOT
   retain Path state and send a ResvTear further upstream because that
   would imply that Path state further downstream had also been
   retained.

7. VPN Connections

   As stated in the addressing section above, the extensions in this
   document are designed to be compatible with the support of VPNs.
   Since the core network may be some technology other than GMPLS, no
   mandatory means of mapping core connections to access connections is
   specified.  However, when GMPLS is used for the core network, it is
   RECOMMENDED that the following procedure based on [MPLS-HIER] is
   followed.

   The VPN connection is modeled as being three hops.  One for each
   access link and one hop across the core network.

   The VPN connection is established using a two-step procedure.  When a
   Path message is received at a core-node on an interface that is part
   of a VPN, the Path message is held until a core connection is
   established.

   The connection across the core is setup as a separate signaling
   exchange between the core-nodes, using the address space of the core
   nodes.  While this exchange is in progress, the original Path message
   is held at the ingress core-node.  Once the exchange for the core
   connection is complete, this connection is used in the VPN connection
   as if it were a single link.  This is signaled by including an IF_ID
   RSVP_HOP object (defined in [RFC3473]) using the procedures defined
   in [MPLS-HIER].

   The original Path message is then forwarded within the VPN addressing
   realm to the core-node attached to the destination edge-node.  Many
   ways of accomplishing this are available, including IP and GRE
   tunnels and BGP/MPLS VPNs.  Specifying a particular means is beyond
   the scope of this document.

Swallow, et al.             Standards Track                     [Page 9]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

8. Security Considerations

   The trust model between the core and edge-nodes is different than the
   one described in [RFC3473], as the core is permitted to hide its
   topology from the edge-nodes, and the core is permitted to restrict
   the actions of edge-nodes by filtering out specific RSVP objects.

9. Acknowledgments

   The authors would like to thank Kireeti Kompella, Jonathan Lang,
   Dimitri Papadimitriou, Dimitrios Pendarakis, Bala Rajagopalan, and
   Adrian Farrel for their comments and input.  Thanks for thorough
   final reviews from Loa Andersson and Dimitri Papadimitriou.

   Adrian Farrel edited the last two revisions of this document to
   incorporate comments from Working Group last call and from AD review.

10.  References

10.1. Normative References

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

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

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

   [RFC3209]    Awduche, D., Berger, L., Gan, D., Li, T., Srinivasan,
                V., and G.  Swallow, "RSVP-TE: Extensions to RSVP for
                LSP Tunnels", RFC 3209, December 2001.

10.2. Informational References

   [RFC3031]    Rosen, E., Viswanathan, A., and R.  Callon,
                "Multiprotocol Label Switching Architecture", RFC 3031,
                January 2001.

   [RFC3477]    Kompella, K. and Y.  Rekhter, "Signalling Unnumbered
                Links in Resource ReSerVation Protocol - Traffic
                Engineering (RSVP-TE)", RFC 3477, January 2003.

Swallow, et al.             Standards Track                    [Page 10]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

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

   [EXPLICIT]   Berger, L., "GMPLS Signaling Procedure for Egress
                Control", RFC 4003, February 2005.

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

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

   [G.8080]     ITU-T Rec.  G.8080/Y.1304, "Architecture for the
                Automatically Switched Optical Network (ASON)," November
                2001 (and Revision, January 2003).  For information on
                the availability of this document, please see
                http://www.itu.int.

Swallow, et al.             Standards Track                    [Page 11]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

Authors' Addresses

   George Swallow
   Cisco Systems, Inc.
   1414 Massachusetts Ave,
   Boxborough, MA 01719

   Phone: +1 978 936 1398
   EMail: swallow@cisco.com

   John Drake
   Boeing Satellite Systems
   2300 East Imperial Highway
   El Segundo, CA 90245

   Phone: +1 412 370-3108
   EMail: John.E.Drake2@boeing.com

   Hirokazu Ishimatsu
   G1M Co., Ltd.
   Nishinippori Start up Office 214,
   5-37-5 Nishinippori, Arakawaku,
   Tokyo 116-0013, Japan

   Phone: +81 3 3891 8320
   EMail: hirokazu.ishimatsu@g1m.jp

   Yakov Rekhter
   Juniper Networks, Inc.

   EMail: yakov@juniper.net

Swallow, et al.             Standards Track                    [Page 12]

RFC 4208         RSVP-TE Support for the Overlay Model      October 2005

Full Copyright Statement

   Copyright (C) The Internet Society (2005).

   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 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.

Acknowledgement

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

Swallow, et al.             Standards Track                    [Page 13]