Network Working Group                                  I. Minei (Editor)
Internet-Draft                                               K. Kompella
Intended status: Standards Track                        Juniper Networks
Expires: December 3, 2006                           I. Wijnands (Editor)
                                                               B. Thomas
                                                     Cisco Systems, Inc.
                                                               June 2006


   Label Distribution Protocol Extensions for Point-to-Multipoint and
             Multipoint-to-Multipoint Label Switched Paths
                      draft-ietf-mpls-ldp-p2mp-02

Status of this Memo

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

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

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

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

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

   This Internet-Draft will expire on December 3, 2006.

Copyright Notice

   Copyright (C) The Internet Society (2006).

Abstract

   This document describes extensions to the Label Distribution Protocol
   (LDP) for the setup of point to multi-point (P2MP) and multipoint-to-
   multipoint (MP2MP) Label Switched Paths (LSPs) in Multi-Protocol
   Label Switching (MPLS) networks.  The solution relies on LDP without



Minei (Editor), et al.  Expires December 3, 2006                [Page 1]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   requiring a multicast routing protocol in the network.  Protocol
   elements and procedures for this solution are described for building
   such LSPs in a receiver-initiated manner.  There can be various
   applications for P2MP/MP2MP LSPs, for example IP multicast or support
   for multicast in BGP/MPLS L3VPNs.  Specification of how such
   applications can use a LDP signaled P2MP/MP2MP LSP is outside the
   scope of this document.


Table of Contents

   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  3
     1.1.  Conventions used in this document  . . . . . . . . . . . .  3
     1.2.  Terminology  . . . . . . . . . . . . . . . . . . . . . . .  3
   2.  Setting up P2MP LSPs with LDP  . . . . . . . . . . . . . . . .  4
     2.1.  The P2MP FEC Element . . . . . . . . . . . . . . . . . . .  4
     2.2.  The LDP MP Opaque Value Element  . . . . . . . . . . . . .  6
       2.2.1.  The Generic LSP Identifier . . . . . . . . . . . . . .  6
     2.3.  Using the P2MP FEC Element . . . . . . . . . . . . . . . .  7
       2.3.1.  Label Map  . . . . . . . . . . . . . . . . . . . . . .  8
       2.3.2.  Label Withdraw . . . . . . . . . . . . . . . . . . . .  9
   3.  Shared Trees . . . . . . . . . . . . . . . . . . . . . . . . . 10
   4.  Setting up MP2MP LSPs with LDP . . . . . . . . . . . . . . . . 11
     4.1.  The MP2MP downstream and upstream FEC elements.  . . . . . 11
     4.2.  Using the MP2MP FEC elements . . . . . . . . . . . . . . . 11
       4.2.1.  MP2MP Label Map upstream and downstream  . . . . . . . 13
       4.2.2.  MP2MP Label Withdraw . . . . . . . . . . . . . . . . . 15
   5.  Upstream label allocation on Ethernet networks . . . . . . . . 16
   6.  Root node redundancy for MP2MP LSPs  . . . . . . . . . . . . . 16
     6.1.  Root node redundancy procedure . . . . . . . . . . . . . . 16
   7.  Make before break  . . . . . . . . . . . . . . . . . . . . . . 17
     7.1.  Protocol event . . . . . . . . . . . . . . . . . . . . . . 18
   8.  Security Considerations  . . . . . . . . . . . . . . . . . . . 18
   9.  IANA considerations  . . . . . . . . . . . . . . . . . . . . . 18
   10. Acknowledgments  . . . . . . . . . . . . . . . . . . . . . . . 19
   11. Contributing authors . . . . . . . . . . . . . . . . . . . . . 19
   12. References . . . . . . . . . . . . . . . . . . . . . . . . . . 21
     12.1. Normative References . . . . . . . . . . . . . . . . . . . 21
     12.2. Informative References . . . . . . . . . . . . . . . . . . 21
   Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 21
   Intellectual Property and Copyright Statements . . . . . . . . . . 23










Minei (Editor), et al.  Expires December 3, 2006                [Page 2]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


1.  Introduction

   The LDP protocol is described in [1].  It defines mechanisms for
   setting up point-to-point (P2P) and multipoint-to-point (MP2P) LSPs
   in the network.  This document describes extensions to LDP for
   setting up point-to-multipoint (P2MP) and multipoint-to-multipoint
   (MP2MP) LSPs.  These are collectively referred to as multipoint LSPs
   (MP LSPs).  A P2MP LSP allows traffic from a single root (or ingress)
   node to be delivered to a number of leaf (or egress) nodes.  A MP2MP
   LSP allows traffic from multiple ingress nodes to be delivered to
   multiple egress nodes.  Only a single copy of the packet will be sent
   on any link traversed by the MP LSP (see note at end of
   Section 2.3.1).  This is accomplished without the use of a multicast
   protocol in the network.  There can be several MP LSPs rooted at a
   given ingress node, each with its own identifier.

   The solution assumes that the leaf nodes of the MP LSP know the root
   node and identifier of the MP LSP to which they belong.  The
   mechanisms for the distribution of this information are outside the
   scope of this document.  The specification of how an application can
   use a MP LSP signaled by LDP is also outside the scope of this
   document.

   Interested readers may also wish to peruse the requirement draft [4]
   and other documents [8] and [9].

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 RFC 2119 [2].

1.2.  Terminology

   The following terminology is taken from [4].

   P2P LSP:  An LSP that has one Ingress LSR and one Egress LSR.


   P2MP LSP:  An LSP that has one Ingress LSR and one or more Egress
      LSRs.


   MP2P LSP:  A LSP that has one or more Ingress LSRs and one unique
      Egress LSR.






Minei (Editor), et al.  Expires December 3, 2006                [Page 3]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   MP2MP LSP:  A LSP that connects a set of leaf nodes, acting
      indifferently as ingress or egress.


   MP LSP:  A multipoint LSP, either a P2MP or an MP2MP LSP.


   Ingress LSR:  Source of the P2MP LSP, also referred to as root node.


   Egress LSR:  One of potentially many destinations of an LSP, also
      referred to as leaf node in the case of P2MP and MP2MP LSPs.


   Transit LSR:  An LSR that has one or more directly connected
      downstream LSRs.


   Bud LSR:  An LSR that is an egress but also has one or more directly
      connected downstream LSRs.



2.  Setting up P2MP LSPs with LDP

   A P2MP LSP consists of a single root node, zero or more transit nodes
   and one or more leaf nodes.  Leaf nodes initiate P2MP LSP setup and
   tear-down.  Leaf nodes also install forwarding state to deliver the
   traffic received on a P2MP LSP to wherever it needs to go; how this
   is done is outside the scope of this document.  Transit nodes install
   MPLS forwarding state and propagate the P2MP LSP setup (and tear-
   down) toward the root.  The root node installs forwarding state to
   map traffic into the P2MP LSP; how the root node determines which
   traffic should go over the P2MP LSP is outside the scope of this
   document.

   For the setup of a P2MP LSP with LDP, we define one new protocol
   entity, the P2MP FEC Element to be used in the FEC TLV.  The
   description of the P2MP FEC Element follows.

2.1.  The P2MP FEC Element

   The P2MP FEC Element consists of the address of the root of the P2MP
   LSP and an opaque value.  The opaque value consists of one or more
   LDP MP Opaque Value Elements.  The opaque value is unique within the
   context of the root node.  The combination of (Root Node Address,
   Opaque Value) uniquely identifies a P2MP LSP within the MPLS network.




Minei (Editor), et al.  Expires December 3, 2006                [Page 4]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   The P2MP FEC element is encoded as follows:

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |P2MP Type (TBD)|        Address Family         | Address Length|
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                       Root Node Address                       |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |    Opaque Length              |    Opaque Value ...           |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+                               +
      ~                                                               ~
      |                                                               |
      |                               +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
      |                               |
      +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+



   Type:  The type of the P2MP FEC element is to be assigned by IANA.


   Address Family:  Two octet quantity containing a value from ADDRESS
      FAMILY NUMBERS in [3] that encodes the address family for the Root
      LSR Address.


   Address Length:  Length of the Root LSR Address in octets.


   Root Node Address:  A host address encoded according to the Address
      Family field.


   Opaque Length:  The length of the Opaque Value, in octets.


   Opaque Value:  One or more MP Opaque Value elements, uniquely
      identifying the P2MP LSP in the context of the Root Node.  This is
      described in the next section.

   If the Address Family is IPv4, the Address Length MUST be 4; if the
   Address Family is IPv6, the Address Length MUST be 16.  No other
   Address Lengths are defined at present.

   If the Address Length doesn't match the defined length for the
   Address Family, the receiver SHOULD abort processing the message
   containing the FEC Element, and send an "Unknown FEC" Notification



Minei (Editor), et al.  Expires December 3, 2006                [Page 5]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   message to its LDP peer signaling an error.

   If a FEC TLV contains a P2MP FEC Element, the P2MP FEC Element MUST
   be the only FEC Element in the FEC TLV.

2.2.  The LDP MP Opaque Value Element

   The LDP MP Opaque Value Element is used in the P2MP and MP2MP FEC
   elements defined in subsequent sections.  It carries information that
   is meaningful to leaf (and bud) LSRs, but need not be interpreted by
   non-leaf LSRs.

   The LDP MP Opaque Value Element is encoded as follows:

       0                   1                   2                   3
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       | Type(TBD)     | Length                        | Value ...     |
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+               |
       ~                                                               ~
       |                                                               |
       |                               +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
       |                               |

       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+


   Type:  The type of the LDP MP Opaque Value Element is to be assigned
      by IANA.


   Length:  The length of the Value field, in octets.


   Value:  String of Length octets, to be interpreted as specified by
      the Type field.

2.2.1.  The Generic LSP Identifier

   The generic LSP identifier is a type of Opaque Value Element encoded
   as follows:

   Type:  1 (to be assigned by IANA)








Minei (Editor), et al.  Expires December 3, 2006                [Page 6]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   Length:  4


   Value:  A 32bit integer, unique in the context of the root, as
      identified by the root's address.

   This type of opaque value element is recommended when mapping of
   traffic to LSPs is non-algorithmic, and done by means outside LDP.

2.3.  Using the P2MP FEC Element

   This section defines the rules for the processing and propagation of
   the P2MP FEC Element.  The following notation is used in the
   processing rules:

   1.  P2MP FEC Element <X, Y>: a FEC Element with Root Node Address X
       and Opaque Value Y.


   2.  P2MP Label Map <X, Y, L>: a Label Map message with a FEC TLV with
       a single P2MP FEC Element <X, Y> and Label TLV with label L.


   3.  P2MP Label Withdraw <X, Y, L>: a Label Withdraw message with a
       FEC TLV with a single P2MP FEC Element <X, Y> and Label TLV with
       label L.


   4.  P2MP LSP <X, Y> (or simply <X, Y>): a P2MP LSP with Root Node
       Address X and Opaque Value Y.

   5.  The notation L' -> {<I1, L1> <I2, L2> ..., <In, Ln>} on LSR X
       means that on receiving a packet with label L', X makes n copies
       of the packet.  For copy i of the packet, X swaps L' with Li and
       sends it out over interface Ii.

   The procedures below are organized by the role which the node plays
   in the P2MP LSP.  Node Z knows that it is a leaf node by a discovery
   process which is outside the scope of this document.  During the
   course of protocol operation, the root node recognizes its role
   because it owns the Root Node Address.  A transit node is any node
   (other than the root node) that receives a P2MP Label Map message
   (i.e., one that has leaf nodes downstream of it).

   Note that a transit node (and indeed the root node) may also be a
   leaf node.





Minei (Editor), et al.  Expires December 3, 2006                [Page 7]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


2.3.1.  Label Map

   The following lists procedures for generating and processing P2MP
   Label Map messages for nodes that participate in a P2MP LSP.  An LSR
   should apply those procedures that apply to it, based on its role in
   the P2MP LSP.

   For the approach described here we use downstream assigned labels.
   On Ethernet networks this may be less optimal, see Section 5.

2.3.1.1.  Determining one's 'upstream LSR'

   A node Z that is part of P2MP LSP <X, Y> determines the LDP peer U
   which lies on the best path from Z to the root node X. If there are
   more than one such LDP peers, only one of them is picked.  U is Z's
   "Upstream LSR" for <X, Y>.

   When there are several candidate upstream LSRs, the LSR MAY select
   one upstream LSR using the following procedure:

   1.  The candidate upstream LSRs are numbered from lower to higher IP
       address

   2.  The following hash is performed: H = (Sum Opaque value) modulo N,
       where N is the number of candidate upstream LSRs

   3.  The selected upstream LSR U is the LSR that has the number H.

   This allows for load balancing of a set of LSPs among a set of
   candidate upstream LSRs, while ensuring that on a LAN interface a
   single upstream LSR is selected.

2.3.1.2.  Leaf Operation

   A leaf node Z of P2MP LSP <X, Y> determines its upstream LSR U for
   <X, Y> as per Section 2.3.1.1, allocates a label L, and sends a P2MP
   Label Map <X, Y, L> to U.

2.3.1.3.  Transit Node operation

   Suppose a transit node Z receives a P2MP Label Map <X, Y, L> from LDP
   peer T. Z checks whether it already has state for <X, Y>.  If not, Z
   allocates a label L', and installs state to swap L' with L over
   interface I associated with peer T. Z also determines its upstream
   LSR U for <X, Y> as per Section 2.3.1.1, and sends a P2MP Label Map
   <X, Y, L'> to U.

   If Z already has state for <X, Y>, then Z does not send a Label Map



Minei (Editor), et al.  Expires December 3, 2006                [Page 8]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   message for P2MP LSP <X, Y>.  All that Z needs to do in this case is
   update its forwarding state.  Assuming its old forwarding state was
   L'-> {<I1, L1> <I2, L2> ..., <In, Ln>}, its new forwarding state
   becomes L'-> {<I1, L1> <I2, L2> ..., <In, Ln>, <I, L>}.

2.3.1.4.  Root Node Operation

   Suppose the root node Z receives a P2MP Label Map <X, Y, L> from peer
   T. Z checks whether it already has forwarding state for <X, Y>.  If
   not, Z creates forwarding state to push label L onto the traffic that
   Z wants to forward over the P2MP LSP (how this traffic is determined
   is outside the scope of this document).

   If Z already has forwarding state for <X, Y>, then Z adds "push label
   L, send over interface I" to the nexthop, where I is the interface
   associated with peer T.

2.3.2.  Label Withdraw

   The following lists procedures for generating and processing P2MP
   Label Withdraw messages for nodes that participate in a P2MP LSP.  An
   LSR should apply those procedures that apply to it, based on its role
   in the P2MP LSP.

2.3.2.1.  Leaf Operation

   If a leaf node Z discovers (by means outside the scope of this
   document) that it is no longer a leaf of the P2MP LSP, it SHOULD send
   a Label Withdraw <X, Y, L> to its upstream LSR U for <X, Y>, where L
   is the label it had previously advertised to U for <X, Y>.

2.3.2.2.  Transit Node Operation

   If a transit node Z receives a Label Withdraw message <X, Y, L> from
   a node W, it deletes label L from its forwarding state, and sends a
   Label Release message with label L to W.

   If deleting L from Z's forwarding state for P2MP LSP <X, Y> results
   in no state remaining for <X, Y>, then Z propagates the Label
   Withdraw for <X, Y>, to its upstream T, by sending a Label Withdraw
   <X, Y, L1> where L1 is the label Z had previously advertised to T for
   <X, Y>.

2.3.2.3.  Root Node Operation

   The procedure when the root node of a P2MP LSP receives a Label
   Withdraw message are the same as for transit nodes, except that it
   would not propagate the Label Withdraw upstream (as it has no



Minei (Editor), et al.  Expires December 3, 2006                [Page 9]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   upstream).

2.3.2.4.  Upstream LSR change

   If, for a given node Z participating in a P2MP LSP <X, Y>, the
   upstream LSR changes, say from U to U', then Z MUST update its
   forwarding state by deleting the state for label L, allocating a new
   label, L', for <X,Y>, and installing the forwarding state for L'.  In
   addition Z MUST send a Label Map <X, Y, L'> to U' and send a Label
   Withdraw <X, Y, L> to U.


3.  Shared Trees

   The mechanism described above shows how to build a tree with a single
   root and multiple leaves, i.e., a P2MP LSP.  One can use essentially
   the same mechanism to build Shared Trees with LDP.  A Shared Tree can
   be used by a group of routers that want to multicast traffic among
   themselves, i.e., each node is both a root node (when it sources
   traffic) and a leaf node (when any other member of the group sources
   traffic).  A Shared Tree offers similar functionality to a MP2MP LSP,
   but the underlying multicasting mechanism uses a P2MP LSP.  One
   example where a Shared Tree is useful is video-conferencing.  Another
   is Virtual Private LAN Service (VPLS) [7], where for some types of
   traffic, each device participating in a VPLS must send packets to
   every other device in that VPLS.

   One way to build a Shared Tree is to build an LDP P2MP LSP rooted at
   a common point, the Shared Root (SR), and whose leaves are all the
   members of the group.  Each member of the Shared Tree unicasts
   traffic to the SR (using, for example, the MP2P LSP created by the
   unicast LDP FEC advertised by the SR); the SR then splices this
   traffic into the LDP P2MP LSP.  The SR may be (but need not be) a
   member of the multicast group.

   A major advantage of this approach is that no further protocol
   mechanisms beyond the one already described are needed to set up a
   Shared Tree.  Furthermore, a Shared Tree is very efficient in terms
   of the multicast state in the network, and is reasonably efficient in
   terms of the bandwidth required to send traffic.

   A property of this approach is that a sender will receive its own
   packets as part of the multicast; thus a sender must be prepared to
   recognize and discard packets that it itself has sent.  For a number
   of applications (for example, VPLS), this requirement is easy to
   meet.  Another consideration is the various techniques that can be
   used to splice unicast LDP MP2P LSPs to the LDP P2MP LSP; these will
   be described in a later revision.



Minei (Editor), et al.  Expires December 3, 2006               [Page 10]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


4.  Setting up MP2MP LSPs with LDP

   An MP2MP LSP is much like a P2MP LSP in that it consists of a single
   root node, zero or more transit nodes and one or more leaf LSRs
   acting equally as Ingress or Egress LSR.  A leaf node participates in
   the setup of an MP2MP LSP by establishing both a downstream LSP,
   which is much like a P2MP LSP from the root, and an upstream LSP
   which is used to send traffic toward the root and other leaf nodes.
   Transit nodes support the setup by propagating the upstream and
   downstream LSP setup toward the root and installing the necessary
   MPLS forwarding state.  The transmission of packets from the root
   node of a MP2MP LSP to the receivers is identical to that for a P2MP
   LSP.  Traffic from a leaf node follows the upstream LSP toward the
   root node and branches downward along the downstream LSP as required
   to reach other leaf nodes.  Mapping traffic to the MP2MP LSP may
   happen at any leaf node.  How that mapping is established is outside
   the scope of this document.

   Due to how a MP2MP LSP is built a leaf LSR that is sending packets on
   the MP2MP LSP does not receive its own packets.  There is also no
   additional mechanism needed on the root or transit LSR to match
   upstream traffic to the downstream forwarding state.  Packets that
   are forwarded over a MP2MP LSP will not traverse a link more than
   once, with the exception of LAN links which are discussed in
   Section 4.2.1

   For the setup of a MP2MP LSP with LDP we define 2 new protocol
   entities, the MP2MP downstream FEC and upstream FEC element.  Both
   elements will be used in the FEC TLV.  The description of the MP2MP
   elements follow.

4.1.  The MP2MP downstream and upstream FEC elements.

   The structure, encoding and error handling for the MP2MP downstream
   and upstream FEC elements are the same as for the P2MP FEC element
   described in Section 2.1.  The difference is that two new FEC types
   are used: MP2MP downstream type (TBD) and MP2MP upstream type (TBD).

   If a FEC TLV contains an MP2MP FEC Element, the MP2MP FEC Element
   MUST be the only FEC Element in the FEC TLV.

4.2.  Using the MP2MP FEC elements

   This section defines the rules for the processing and propagation of
   the MP2MP FEC elements.  The following notation is used in the
   processing rules:





Minei (Editor), et al.  Expires December 3, 2006               [Page 11]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   1.  MP2MP downstream LSP <X, Y> (or simply downstream <X, Y>): an
       MP2MP LSP downstream path with root node address X and opaque
       value Y.


   2.  MP2MP upstream LSP <X, Y, D> (or simply upstream <X, Y, D>): a
       MP2MP LSP upstream path for downstream node D with root node
       address X and opaque value Y.


   3.  MP2MP downstream FEC element <X, Y>: a FEC element with root node
       address X and opaque value Y used for a downstream MP2MP LSP.


   4.  MP2MP upstream FEC element <X, Y>: a FEC element with root node
       address X and opaque value Y used for an upstream MP2MP LSP.


   5.  MP2MP Label Map downstream <X, Y, L>: A Label Map message with a
       FEC TLV with a single MP2MP downstream FEC element <X, Y> and
       label TLV with label L.


   6.  MP2MP Label Map upstream <X, Y, Lu>: A Label Map message with a
       FEC TLV with a single MP2MP upstream FEC element <X, Y> and label
       TLV with label Lu.


   7.  MP2MP Label Withdraw downstream <X, Y, L>: a Label Withdraw
       message with a FEC TLV with a single MP2MP downstream FEC element
       <X, Y> and label TLV with label L.


   8.  MP2MP Label Withdraw upstream <X, Y, Lu>: a Label Withdraw
       message with a FEC TLV with a single MP2MP upstream FEC element
       <X, Y> and label TLV with label Lu.

   The procedures below are organized by the role which the node plays
   in the MP2MP LSP.  Node Z knows that it is a leaf node by a discovery
   process which is outside the scope of this document.  During the
   course of the protocol operation, the root node recognizes its role
   because it owns the root node address.  A transit node is any node
   (other then the root node) that receives a MP2MP Label Map message
   (i.e., one that has leaf nodes downstream of it).

   Note that a transit node (and indeed the root node) may also be a
   leaf node and the root node does not have to be an ingress LSR or
   leaf of the MP2MP LSP.



Minei (Editor), et al.  Expires December 3, 2006               [Page 12]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


4.2.1.  MP2MP Label Map upstream and downstream

   The following lists procedures for generating and processing MP2MP
   Label Map messages for nodes that participate in a MP2MP LSP.  An LSR
   should apply those procedures that apply to it, based on its role in
   the MP2MP LSP.

   For the approach described here if there are several receivers for a
   MP2MP LSP on a LAN, packets are replicated over the LAN.  This may
   not be optimal; optimizing this case is for further study, see [5].

4.2.1.1.  Determining one's upstream MP2MP LSR

   Determining the upstream LDP peer U for a MP2MP LSP <X, Y> follows
   the procedure for a P2MP LSP described in Section 2.3.1.1.

4.2.1.2.  Determining one's downstream MP2MP LSR

   A LDP peer U which receives a MP2MP Label Map downstream from a LDP
   peer D will treat D as downstream MP2MP LSR.

4.2.1.3.  MP2MP leaf node operation

   A leaf node Z of a MP2MP LSP <X, Y> determines its upstream LSR U for
   <X, Y> as per Section 4.2.1.1, allocates a label L, and sends a MP2MP
   Label Map downstream <X, Y, L> to U.

   Leaf node Z expects an MP2MP Label Map upstream <X, Y, Lu> from node
   U in response to the MP2MP Label Map downstream it sent to node U. Z
   checks whether it already has forwarding state for upstream <X, Y>.
   If not, Z creates forwarding state to push label Lu onto the traffic
   that Z wants to forward over the MP2MP LSP.  How it determines what
   traffic to forward on this MP2MP LSP is outside the scope of this
   document.

4.2.1.4.  MP2MP transit node operation

   When node Z receives a MP2MP Label Map downstream <X, Y, L> from peer
   D associated with interface I, it checks whether it has forwarding
   state for downstream <X, Y>.  If not, Z allocates a label L' and
   installs downstream forwarding state to swap label L' with label L
   over interface I. Z also determines its upstream LSR U for <X, Y> as
   per Section 4.2.1.1, and sends a MP2MP Label Map downstream <X, Y,
   L'> to U.

   If Z already has forwarding state for downstream <X, Y>, all that Z
   needs to do is update its forwarding state.  Assuming its old
   forwarding state was L'-> {<I1, L1> <I2, L2> ..., <In, Ln>}, its new



Minei (Editor), et al.  Expires December 3, 2006               [Page 13]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   forwarding state becomes L'-> {<I1, L1> <I2, L2> ..., <In, Ln>, <I,
   L>}.

   Node Z checks whether it already has forwarding state upstream <X, Y,
   D>.  If it does, then no further action needs to happen.  If it does
   not, it allocates a label Lu and creates a new label swap for Lu from
   the label swap(s) from the forwarding state downstream <X, Y>,
   omitting the swap on interface I for node D. This allows upstream
   traffic to follow the MP2MP tree down to other node(s) except the
   node from which Z received the MP2MP Label Map downstream <X, Y, L>.
   Node Z determines the downstream MP2MP LSR as per Section 4.2.1.2,
   and sends a MP2MP Label Map upstream <X, Y, Lu> to node D.

   Transit node Z will also receive a MP2MP Label Map upstream <X, Y,
   Lu> in response to the MP2MP Label Map downstream sent to node U
   associated with interface Iu.  Node Z will add label swap Lu over
   interface Iu to the forwarding state upstream <X, Y, D>.  This allows
   packets to go up the tree towards the root node.

4.2.1.5.  MP2MP root node operation

4.2.1.5.1.  Root node is also a leaf

   Suppose root/leaf node Z receives a MP2MP Label Map downstream <X, Y,
   L> from node D associated with interface I. Z checks whether it
   already has forwarding state downstream <X, Y>.  If not, Z creates
   forwarding state for downstream to push label L on traffic that Z
   wants to forward down the MP2MP LSP.  How it determines what traffic
   to forward on this MP2MP LSP is outside the scope of this document.
   If Z already has forwarding state for downstream <X, Y>, then Z will
   add the label push for L over interface I to it.

   Node Z checks if it has forwarding state for upstream <X, Y, D>.  If
   not, Z allocates a label Lu and creates upstream forwarding state to
   push Lu with the label push(s) from the forwarding state downstream
   <X, Y>, except the push on interface I for node D. This allows
   upstream traffic to go down the MP2MP to other node(s), except the
   node from which the traffic was received.  Node Z determines the
   downstream MP2MP LSR as per section Section 4.2.1.2, and sends a
   MP2MP Label Map upstream <X, Y, Lu> to node D. Since Z is the root of
   the tree Z will not send a MP2MP downstream map and will not receive
   a MP2MP upstream map.

4.2.1.5.2.  Root node is not a leaf

   Suppose the root node Z receives a MP2MP Label Map dowbstream <X, Y,
   L> from node D associated with interface I. Z checks whether it
   already has forwarding state for downstream <X, Y>.  If not, Z



Minei (Editor), et al.  Expires December 3, 2006               [Page 14]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   creates downstream forwarding state and installs a outgoing label L
   over interface I. If Z already has forwarding state for downstream
   <X, Y>, then Z will add label L over interface I to the existing
   state.

   Node Z checks if it has forwarding state for upstream <X, Y, D>.  If
   not, Z allocates a label Lu and creates forwarding state to swap Lu
   with the label swap(s) from the forwarding state downstream <X, Y>,
   except the swap for node D. This allows upstream traffic to go down
   the MP2MP to other node(s), except the node is was received from.
   Root node Z determines the downstream MP2MP LSR D as per
   Section 4.2.1.2, and sends a MP2MP Label Map upstream <X, Y, Lu> to
   it.  Since Z is the root of the tree Z will not send a MP2MP
   downstream map and will not receive a MP2MP upstream map.

4.2.2.  MP2MP Label Withdraw

   The following lists procedures for generating and processing MP2MP
   Label Withdraw messages for nodes that participate in a MP2MP LSP.
   An LSR should apply those procedures that apply to it, based on its
   role in the MP2MP LSP.

4.2.2.1.  MP2MP leaf operation

   If a leaf node Z discovers (by means outside the scope of this
   document) that it is no longer a leaf of the MP2MP LSP, it SHOULD
   send a downstream Label Withdraw <X, Y, L> to its upstream LSR U for
   <X, Y>, where L is the label it had previously advertised to U for
   <X,Y>.

   Leaf node Z expects the upstream router U to respond by sending a
   downstream label release for L and a upstream Label Withdraw for <X,
   Y, Lu> to remove Lu from the upstream state.  Node Z will remove
   label Lu from its upstream state and send a label release message
   with label Lu to U.

4.2.2.2.  MP2MP transit node operation

   If a transit node Z receives a downstream label withdraw message <X,
   Y, L> from node D, it deletes label L from its forwarding state
   downstream <X, Y> and from all its upstream states for <X, Y>.  Node
   Z sends a label release message with label L to D. Since node D is no
   longer part of the downstream forwarding state, Z cleans up the
   forwarding state upstream <X, Y, D> and sends a upstream Label
   Withdraw for <X, Y, Lu> to D.

   If deleting L from Z's forwarding state for downstream <X, Y> results
   in no state remaining for <X, Y>, then Z propagates the Label



Minei (Editor), et al.  Expires December 3, 2006               [Page 15]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   Withdraw <X, Y, L> to its upstream node U for <X,Y>.

4.2.2.3.  MP2MP root node operation

   The procedure when the root node of a MP2MP LSP receives a label
   withdraw message is the same as for transit nodes, except that the
   root node would not propagate the Label Withdraw upstream (as it has
   no upstream).

4.2.2.4.  MP2MP Upstream LSR change

   The procedure for changing the upstream LSR is the same as documented
   in Section 2.3.2.4, except it is applied to MP2MP FECs, using the
   procedures described in Section 4.2.1 through Section 4.2.2.3.


5.  Upstream label allocation on Ethernet networks

   On Ethernet networks the upstream LSR will send a copy of the packet
   to each receiver individually.  If there is more then one receiver on
   the Ethernet we don't take full benefit of the multi-access
   capability of the network.  We may optimize the bandwidth consumption
   on the Ethernet and replication overhead on the upstream LSR by using
   upstream label allocation [5].  Procedures on how to distribute
   upstream labels using LDP is documented in [6].


6.  Root node redundancy for MP2MP LSPs

   MP2MP leaf nodes must use the same root node to setup the MP2MP LSP.
   Otherwise there will be partitioned MP2MP LSP and traffic sourced by
   some leafs is not received by others.  Having a single root node for
   a MP2MP LSP is a single point of failure, which is not preferred.  We
   need a fast and efficient mechanism to recover from a root node
   failure.

6.1.  Root node redundancy procedure

   It is likely that the root node for a MP2MP LSP is defined
   statically.  The root node address may be configured on each leaf
   statically or learned using a dynamic protocol.  How MP2MP leafs
   learn about the root node is out of the scope of this document.  A
   MP2MP LSP is uniquely identified by a opaque value and the root node
   address.  Suppose that for the same opaque value we define two root
   node addresses and we build a tree to each root using the same opaque
   value.  Effectively these will be treated as different MP2MP LSPs in
   the network.  Since all leafs have setup a MP2MP LSP to each one of
   the root nodes for this opaque value, a sending leaf may pick either



Minei (Editor), et al.  Expires December 3, 2006               [Page 16]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   of the two MP2MP LSPs to forward a packet on.  The leaf nodes will
   receive the packet on one of the MP2MP LSPs, the client of the MP2MP
   LSP does not care on which MP2MP LSP the packet was received from, as
   long as they are for the same opaque value.  The sending leaf MUST
   only forward a packet on one MP2MP LSP at a given point in time.  The
   receiving leafs are unable to discard duplicate packets because they
   accept on both LSPs.  Using both these MP2MP LSPs we can implement
   redundancy using the following procedures.

   A sending leaf selects a single root node out of the available roots
   for a given opaque value.  A good strategy MAY be to look at the
   unicast routing table and select a root that is closest according in
   terms of unicast metric.  As soon as the root address of our active
   root disappears from the unicast routing table (or becomes less
   attractive) due to root node or link failure we can select a new best
   root address and start forwarding to it directly.  If multiple root
   nodes have the same unicast metric, the highest root node addresses
   MAY be selected, or we MAY do per session load balancing over the
   root nodes.

   All leafs participating in a MP2MP LSP MUST join to all the available
   root nodes for a given opaque value.  Since the sending leaf may pick
   any MP2MP LSP, it must be prepared to receive on it.

   The advantage of pre-building multiple MP2MP LSPs for a single opaque
   value is that we can converge from a root node failure as fast as the
   unicast routing protocol is able to notify us.  There is no need for
   an additional protocol to advertise to the leaf nodes which root node
   is the active root.  The root selection is a local leaf policy that
   does not need to be coordinated with other leafs.  The disadvantage
   is that we are using more label resources depending on how many root
   nodes are defined.


7.  Make before break

   An upstream LSR is chosen based on the best path to reach the root of
   the MP LSP.  When the best path to reach the root changes it needs to
   choose a new upstream LSR.  Section 2.3.2.4 and Section 4.2.2.4
   describes these procedures.  When the best path to the root changes
   the LSP may be broken and packet forwarding is interrupted, in that
   case it needs to converge to a new upstream LSR ASAP.  There are also
   scenarios where the best path changed, but the LSP is still
   forwarding packets.  That happens when links come up or routing
   metrics are changed.  In that case it would like to build the new LSP
   before it breaks the old LSP to minimize the traffic interruption.
   The approuch described below deals with both scenarios and does not
   require LDP to know which of the events above caused the upstream



Minei (Editor), et al.  Expires December 3, 2006               [Page 17]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   router to change.  The approuch below is an optional extention to the
   MP LSP building procedures described in this draft.

7.1.  Protocol event

   An approach is to use additional signaling in LDP.  Suppose a
   downstream LSR-D is changing to a new upstream LSR-U for FEC-A, this
   LSR-U may already be forwarding packets for this FEC-A.  Based on the
   existence of state for FEC-A, LSR-U will send a notification to the
   LSR-D to initiate the switchover.  The assumption is that if our
   upstream LSR-U has state for the FEC-A and it has received a
   notification from its upstream router, then this LSR is forwarding
   packets for this FEC-A and it can send a notification back to
   initiate the switchover.  You could say there is an explicit
   notification to tell the LSR it became part of the tree identified by
   FEC-A.  LSR-D can be in 3 different states.

   1.  There no state for a given FEC-A.

   2.  State for FEC-A has just been created and is waiting for
       notification.

   3.  State for FEC-A exists and notification was received.

   Suppose LSR-D sends a label mapping for FEC-A to LSR-U.  LSR-U must
   only reply with a notification to LSR-D if it is in state #3 as
   described above.  If LSR-U is in state 1 or 2, it should remember it
   has received a label mapping from LSR-D which is waiting for a
   notification.  As soon as LSR-U received a notification from its
   upstream LSR it can move to state #3 and trigger notifications to its
   downstream LSR's that requested it.  More details will be added in
   the next revision of the draft.


8.  Security Considerations

   The same security considerations apply as for the base LDP
   specification, as described in [1].


9.  IANA considerations

   This document creates a new name space (the LDP MP Opaque Value
   Element type) that is to be managed by IANA.  Also, this document
   requires allocation of three new LDP FEC element types: the P2MP
   type, the MP2MP-up and the MP2MP-down types.





Minei (Editor), et al.  Expires December 3, 2006               [Page 18]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


10.  Acknowledgments

   The authors would like to thank the following individuals for their
   review and contribution: Nischal Sheth, Yakov Rekhter, Rahul
   Aggarwal, Arjen Boers, Eric Rosen, Nidhi Bhaskar, Toerless Eckert and
   George Swallow.


11.  Contributing authors

   Below is a list of the contributing authors in alphabetical order:

   Shane Amante
   Level 3 Communications, LLC
   1025 Eldorado Blvd
   Broomfield, CO 80021
   US
   Email: Shane.Amante@Level3.com


   Luyuan Fang
   AT&T
   200 Laurel Avenue, Room C2-3B35
   Middletown, NJ 07748
   US
   Email: luyuanfang@att.com


   Hitoshi Fukuda
   NTT Communications Corporation
   1-1-6, Uchisaiwai-cho, Chiyoda-ku
   Tokyo 100-8019,
   Japan
   Email: hitoshi.fukuda@ntt.com


   Yuji Kamite
   NTT Communications Corporation
   Tokyo Opera City Tower
   3-20-2 Nishi Shinjuku, Shinjuku-ku,
   Tokyo 163-1421,
   Japan
   Email: y.kamite@ntt.com








Minei (Editor), et al.  Expires December 3, 2006               [Page 19]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   Kireeti Kompella
   Juniper Networks
   1194 N. Mathilda Ave.
   Sunnyvale, CA 94089
   US
   Email: kireeti@juniper.net


   Ina Minei
   Juniper Networks
   1194 N. Mathilda Ave.
   Sunnyvale, CA  94089
   US
   Email: ina@juniper.net


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


   Bob Thomas
   Cisco Systems, Inc.
   300 Beaver Brook Road
   Boxborough, MA, 01719
   E-mail: rhthomas@cisco.com


   Lei Wang
   Telenor
   Snaroyveien 30
   Fornebu 1331
   Norway
   Email: lei.wang@telenor.com


   IJsbrand Wijnands
   Cisco Systems, Inc.
   De kleetlaan 6a
   1831 Diegem
   Belgium
   E-mail: ice@cisco.com


12.  References



Minei (Editor), et al.  Expires December 3, 2006               [Page 20]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


12.1.  Normative References

   [1]  Andersson, L., Doolan, P., Feldman, N., Fredette, A., and B.
        Thomas, "LDP Specification", RFC 3036, January 2001.

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

   [3]  Reynolds, J. and J. Postel, "Assigned Numbers", RFC 1700,
        October 1994.

   [4]  Roux, J., "Requirements for point-to-multipoint extensions to
        the Label Distribution  Protocol",
        draft-leroux-mpls-mp-ldp-reqs-03 (work in progress),
        February 2006.

   [5]  Aggarwal, R., "MPLS Upstream Label Assignment and Context
        Specific Label Space", draft-raggarwa-mpls-upstream-label-01
        (work in progress), October 2005.

   [6]  Aggarwal, R. and J. Roux, "MPLS Upstream Label Assignment for
        RSVP-TE and LDP", draft-raggarwa-mpls-rsvp-ldp-upstream-00 (work
        in progress), July 2005.

12.2.  Informative References

   [7]  Andersson, L. and E. Rosen, "Framework for Layer 2 Virtual
        Private Networks (L2VPNs)", draft-ietf-l2vpn-l2-framework-05
        (work in progress), June 2004.

   [8]  Aggarwal, R., "Extensions to RSVP-TE for Point-to-Multipoint TE
        LSPs", draft-ietf-mpls-rsvp-te-p2mp-06 (work in progress),
        August 2006.

   [9]  Rosen, E. and R. Aggarwal, "Multicast in MPLS/BGP IP VPNs",
        draft-ietf-l3vpn-2547bis-mcast-02 (work in progress), June 2006.


Authors' Addresses

   Ina Minei
   Juniper Networks
   1194 N. Mathilda Ave.
   Sunnyvale, CA  94089
   US

   Email: ina@juniper.net




Minei (Editor), et al.  Expires December 3, 2006               [Page 21]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


   Kireeti Kompella
   Juniper Networks
   1194 N. Mathilda Ave.
   Sunnyvale, CA  94089
   US

   Email: kireeti@juniper.net


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

   Email: ice@cisco.com


   Bob Thomas
   Cisco Systems, Inc.
   300 Beaver Brook Road
   Boxborough  01719
   US

   Email: rhthomas@cisco.com


























Minei (Editor), et al.  Expires December 3, 2006               [Page 22]


Internet-Draft      P2MP and MP2MP LSP Setup with LDP          June 2006


Full Copyright Statement

   Copyright (C) The Internet Society (2006).

   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.


Acknowledgment

   Funding for the RFC Editor function is provided by the IETF
   Administrative Support Activity (IASA).





Minei (Editor), et al.  Expires December 3, 2006               [Page 23]