Internet Engineering Task Force (IETF)                      M. Napierala
Request for Comments: 7060                                          AT&T
Category: Standards Track                                       E. Rosen
ISSN: 2070-1721                                             IJ. Wijnands
                                                     Cisco Systems, Inc.
                                                           November 2013


        Using LDP Multipoint Extensions on Targeted LDP Sessions

Abstract

   Label Distribution Protocol (LDP) can be used to set up Point-to-
   Multipoint (P2MP) and Multipoint-to-Multipoint (MP2MP) Label Switched
   Paths.  However, the specification for the Multipoint Extensions to
   LDP presupposes that the two endpoints of an LDP session are directly
   connected.  The LDP base specification allows for the case where the
   two endpoints of an LDP session are not directly connected; such a
   session is known as a "Targeted LDP" session.  This document provides
   the specification for using the LDP Multipoint Extensions over a
   Targeted LDP session.

Status of This Memo

   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc7060.

Copyright Notice

   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.

Table of Contents

   1. Introduction ....................................................2
   2. Targeted mLDP and the Upstream LSR ..............................3
      2.1. Selecting the Upstream LSR .................................3
      2.2. Sending Data from U to D ...................................4
   3. Applicability of Targeted mLDP ..................................4
   4. LDP Capabilities ................................................5
   5. Targeted mLDP with Unicast Replication ..........................5
   6. Targeted mLDP with Multicast Tunneling ..........................6
   7. Security Considerations .........................................8
   8. Acknowledgments .................................................8
   9. Normative References ............................................8

1.  Introduction

   Label Distribution Protocol (LDP) extensions for setting up Point-to-
   Multipoint (P2MP) Label Switched Paths (LSPs) and Multipoint-to-
   Multipoint (MP2MP) LSPs are specified in [mLDP].  This set of
   extensions is generally known as "Multipoint LDP" (mLDP).

   A pair of Label Switched Routers (LSRs) that are the endpoints of an
   LDP session are considered to be "LDP peers".  When a pair of LDP
   peers are "directly connected" (e.g., they are connected by a layer 2
   medium or are otherwise considered to be neighbors by the network's
   interior routing protocol), the LDP session is said to be a "directly
   connected" LDP session.  When the pair of LDP peers are not directly
   connected, the session between them is said to be a "Targeted" LDP
   session.

   The base specification for mLDP does not explicitly cover the case
   where the LDP multipoint extensions are used over a Targeted LDP
   session.  This document provides that specification.

   We will use the term "Multipoint" to mean "either P2MP or MP2MP".

   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.  Targeted mLDP and the Upstream LSR

2.1.  Selecting the Upstream LSR

   In mLDP, a multipoint LSP (MP-LSP) has a unique identifier that is an
   ordered pair of the form <root, opaque value>.  The first element of
   the ordered pair is the IP address of the MP-LSP's "root node".  The
   second element of the ordered pair is an identifier that is unique in
   the context of the root node.

   If LSR D is setting up the MP-LSP <R, X>, D must determine the
   "upstream LSR" for <R, X>.  In [mLDP], the upstream LSR for <R, X>,
   U, is defined to be the "next hop" on D's path to R, and "next hop"
   is tacitly assumed to mean "IGP next hop".  It is thus assumed that
   there is a direct LDP session between D and U.  In this
   specification, we extend the notion of "upstream LSR" to cover the
   following cases:

      -  U is the "BGP next hop" on D's path to R, where U and D are not
         necessarily IGP neighbors, and where there is a Targeted LDP
         session between U and D.  In this case, we allow D to select U
         as the "upstream LSR" for <R,X>.

      -  If the "next-hop interface" on D's path to R is an RSVP Traffic
         Engineering (RSVP-TE) P2P tunnel whose remote endpoint is U,
         and if there is known to be an RSVP-TE P2P tunnel from U to D,
         and if there is a Targeted LDP session between U and D, then we
         allow D to select U as the "upstream LSR" for <R,X>.  This is
         useful when D and U are part of a network area that is fully
         meshed via RSVP-TE P2P tunnels.

   The particular method used to select an "upstream LSR" is determined
   by the Service Provider (SP) and must be made known a priori (i.e.,
   by provisioning) to all the LSRs involved.

   Other methods than the two specified above MAY be used; however, the
   specification of other methods is outside the scope of this document.

2.2.  Sending Data from U to D

   By using Targeted mLDP, we can construct an MP-LSP <R,X> containing
   an LSR U, where U has one or more downstream LSR neighbors (D1, ...,
   Dn) to which it is not directly connected.  In order for a data
   packet to travel along this MP-LSP, U must have some way of
   transmitting the packet to D1, ..., Dn.  We will cover two methods of
   transmission:

      -  Unicast Replication

         In this method, U creates n copies of the packet and unicasts
         each copy to exactly one of D1, ..., Dn.

      -  Multicast Tunneling

         In this method, U becomes the root node of a multicast tunnel,
         with D1, ..., Dn as leaf nodes.  When a packet traveling along
         the MP-LSP <R,X> arrives at U, U transmits it through the
         multicast tunnel, and as a result it arrives at D1, ..., Dn.

         When this method is used, it may be desirable to carry traffic
         of multiple MP-LSPs through a single multicast tunnel.  We
         specify procedures that allow for the proper demultiplexing of
         the MP-LSPs at the leaf nodes of the multicast tunnel.  We do
         not assume that all the leaf nodes of the tunnel are on all the
         MP-LSPs traveling through the tunnel; thus, some of the tunnel
         leaf nodes may need to discard some of the packets received
         through the tunnel.  For example, suppose MP-LSP <R1,X1>
         contains node U with downstream LSRs D1 and D2, while MP-LSP
         <R2,X2> contains node U with downstream LSRs D2 and D3.
         Suppose also that there is a multicast tunnel with U as root
         and with D1, D2, and D3 as leaf nodes.  U can aggregate both
         MP-LSPs in this one tunnel.  However, D1 will have to discard
         packets that are traveling on <R2,X1>, while D3 will have to
         discard packets that are traveling on <R1,X2>.

3.  Applicability of Targeted mLDP

   When LSR D is setting up MP-LSP <R,X>, it MUST NOT use Targeted mLDP
   unless D implements a procedure that can select an LSR U that is a
   Targeted mLDP peer of D as the "upstream LSR" for <R,X>.  See Section
   2.1.

   Whether D uses Targeted mLDP when this condition holds is determined
   by provisioning or by other methods that are outside the scope of
   this specification.

   When Targeted mLDP is used, the choice between unicast replication
   and multicast tunneling is determined by provisioning or by other
   methods that are outside the scope of this specification.  It is
   presupposed that all nodes will have a priori knowledge of whether to
   use unicast replication or to use multicast tunneling.  If the
   latter, it is presupposed that all nodes will have a priori knowledge
   of the type of multicast tunneling to use.

4.  LDP Capabilities

   Per [mLDP], any LSR that needs to set up an MP-LSP must support the
   procedures of [LDP-CAP], and in particular must send and receive the
   P2MP Capability and/or the MP2MP Capability.  This specification does
   not define any new capabilities; the advertisement of the P2MP and/or
   MP2MP Capabilities on a Targeted LDP session means that the
   advertising LSR is capable of following the procedures set forth in
   this document.

   Some of the procedures described in this document require the use of
   upstream-assigned labels [LDP-UP].  In order to use upstream-assigned
   labels as part of Targeted mLDP, an LSR must advertise the LDP
   Upstream-Assigned Label Capability [LDP-UP] on the Targeted LDP
   session.

5.  Targeted mLDP with Unicast Replication

   When unicast replication is used, the mLDP procedures are exactly the
   same as described in [mLDP], with the following exception.  If LSR D
   is setting up MP-LSP <R,X>, its "upstream LSR" is selected according
   to the procedures of Section 2.1, and is not necessarily the "IGP
   next hop" on D's path to R.

   Suppose that LSRs D1 and D2 are both setting up the P2MP MP-LSP
   <R,X>, and that LSR U is the upstream LSR on each of their paths to
   R.  D1 and D2 each binds a label to <R,X> and each uses a Label
   Mapping message to inform U of the label binding.  Suppose D1 has
   assigned label L1 to <R,X> and D2 has assigned label L2 to <R,X>.
   (Note that L1 and L2 could have the same value or different values;
   D1 and D2 do not coordinate their label assignments.)  When U has a
   packet to transmit on the MP-LSP <R,X>, it makes a copy of the
   packet, pushes on label L1, and unicasts the resulting packet to D1.
   It also makes a second copy of the packet, pushes on label L2, and
   then unicasts the resulting packet to D2.

   This procedure also works when the MP-LSP <R,X> is an MP2MP LSP.
   Suppose that in addition to labels L1 and L2 described above, U has
   assigned label L3 for <R,X> traffic received from D1 and label L4 for
   <R,X> traffic received from D2.  When U processes a packet with label

   L3 at the top of its label stack, it knows the packet is from D1, so
   U sends a unicast copy of the packet to D2, after swapping L3 for L2.
   U does not send a copy back to D1.

   Note that all labels used in this procedure are downstream-assigned
   labels.

   The method of unicast is a local matter, outside the scope of this
   specification.  The only requirement is that D1 will receive the copy
   of the packet carrying label L1 and that D1 will process the packet
   by looking up label L1.  (And similarly, D2 must receive the copy of
   the packet carrying label L2 and must process the packet by looking
   up label L2.)

   Note that if the method of unicast is MPLS, U will need to push
   another label on each copy of the packet before transmitting it.
   This label needs to ensure that delivery of the packet to the
   appropriate LSR, D1 or D2.  Use of penultimate-hop popping for that
   label is perfectly legitimate.

6.  Targeted mLDP with Multicast Tunneling

   Suppose that LSRs D1 and D2 are both setting up MP-LSP <R,X> and that
   LSR U is the upstream LSR on each of their paths to R.  Since
   multicast tunneling is being used, when U has a packet to send on
   this MP-LSP, it does not necessarily send two copies, one to D1 and
   one to D2.  It may send only one copy of the packet, which will get
   replicated somewhere downstream in the multicast tunnel.  Therefore,
   the label that gets bound to the MP-LSP must be an upstream-assigned
   label assigned by U.  This requires a change from the procedures of
   [mLDP].  D1 and D2 do not send Label Mapping messages to U; instead,
   they send Label Request messages to U, following the procedures of
   Section 4 of [LDP-UP], asking U to assign a label to the MP-LSP
   <R,X>.  U responds with a Label Mapping message containing an
   upstream-assigned label L (using the procedures specified in
   [LDP-UP]).  As part of the same Label Mapping message, U also sends
   an Interface TLV (as specified in [LDP-UP]) identifying the multicast
   tunnel in which data on the MP-LSP will be carried.  When U transmits
   a packet on this tunnel, it first pushes on the upstream-assigned
   label L and then pushes on the label that corresponds to the
   multicast tunnel.

   If the numerical value L of the upstream-assigned label is the value
   3, defined in [LDP] and [RFC3032] as "Implicit NULL", then the
   specified multicast tunnel will carry only the specified MP-LSP.
   That is, aggregation of multiple MP-LSPs into a single multicast

   tunnel is not being done.  In this case, no upstream-assigned label
   is pushed onto a packet that is transmitted through the multicast
   tunnel.

   Various types of multicast tunnel may be used.  The choice of tunnel
   type is determined by provisioning, or by some other method that is
   outside the scope of this document.  [LDP-UP] specifies encodings
   allowing U to identify an mLDP MP-LSP, and RSVP-TE P2MP LSP, as well
   as other types of multicast tunnel.

   Procedures for tunneling MP2MP LSPs through P2MP or MP2MP LSPs are
   outside the scope of this document.

   If the multicast tunnel is an mLDP MP-LSP or an RSVP-TE P2MP LSP,
   when U transmits a packet on the MP-LSP <R,X>, the upstream-assigned
   label L will be the second label in the label stack.  Penultimate-hop
   popping MUST NOT be done, because the top label provides the context
   in which the second label is to be interpreted.  See [RFC5331].

   When LSR U uses these procedures to inform LSR D that a particular
   MP-LSP is being carried in a particular multicast tunnel, U and D
   MUST take appropriate steps to ensure that the packets U sends into
   this tunnel will be received by D.  The exact steps to take depend on
   the tunnel type.  As long as U is D's upstream LSR for any MP-LSP
   that has been assigned to this tunnel, D must remain joined to the
   tunnel.

   Note that U MAY assign the same multicast tunnel for multiple
   different MP-LSPs.  However, U MUST assign a distinct upstream-
   assigned label to each MP-LSP.  This allows the packets traveling
   through the tunnel to be demultiplexed into the proper MP-LSPs.

   If U has an MP-LSP <R1,X1> with downstream LSRs D1 and D2, and an MP-
   LSP <R2,X2> with downstream LSRs D2 and D3, U may assign both MP-LSPs
   to the same multicast tunnel.  In this case, D3 will receive packets
   traveling on <R1,X1>.  However, the upstream-assigned label carried
   by those packets will not be recognized by D3, hence D3 will discard
   those packets.  Similarly, D1 will discard the <R2,X2> packets.

   This document does not specify any rules for deciding whether to
   aggregate two or more MP-LSPs into a single multicast tunnel.  Such
   rules are outside the scope of this document.

   Except for the procedures explicitly detailed in this document, the
   procedures of [mLDP] and [LDP-UP] apply unchanged.

7.  Security Considerations

   This document raises no new security considerations beyond those
   discussed in [LDP], [LDP-UP], and [RFC5331].

8.  Acknowledgments

   The authors wish to thank Lizhong Jin and Lizhen Bin for their
   comments.

9.  Normative References

   [LDP]      Andersson, L., Ed., Minei, I., Ed., and B. Thomas, Ed.,
              "LDP Specification", RFC 5036, October 2007.

   [LDP-CAP]  Thomas, B., Raza, K., Aggarwal, S., Aggarwal, R., and JL.
              Le Roux, "LDP Capabilities", RFC 5561, July 2009.

   [mLDP]     Wijnands, IJ., Ed., Minei, I., Ed., Kompella, K., and B.
              Thomas, "Label Distribution Protocol Extensions for
              Point-to-Multipoint and Multipoint-to-Multipoint Label
              Switched Paths", RFC 6388, November 2011.

   [LDP-UP]   Aggarwal, R. and JL. Le Roux, "MPLS Upstream Label
              Assignment for LDP", RFC 6389, November 2011.

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

   [RFC3032]  Rosen, E., Tappan, D., Fedorkow, G., Rekhter, Y.,
              Farinacci, D., Li, T., and A. Conta, "MPLS Label Stack
              Encoding", RFC 3032, January 2001.

   [RFC5331]  Aggarwal, R., Rekhter, Y., and E. Rosen, "MPLS Upstream
              Label Assignment and Context-Specific Label Space", RFC
              5331, August 2008.

Authors' Addresses

   Maria Napierala
   AT&T Labs
   200 Laurel Avenue
   Middletown, NJ 07748
   USA

   EMail: mnapierala@att.com


   Eric C. Rosen
   Cisco Systems, Inc.
   1414 Massachusetts Avenue
   Boxborough, MA, 01719
   USA

   EMail: erosen@cisco.com


   IJsbrand Wijnands
   Cisco Systems, Inc.
   De kleetlaan 6a Diegem 1831
   Belgium

   EMail: ice@cisco.com