tech-invite   World Map     

IETF     RFCs     Groups     SIP     ABNFs    |    3GPP     Specs     Glossaries     Architecture     IMS     UICC    |    search     info

RFC 7716

Proposed STD
Pages: 22
Top     in Index     Prev     Next
in Group Index     Prev in Group     Next in Group     Group: BESS

Global Table Multicast with BGP Multicast VPN (BGP-MVPN) Procedures

 


Top       ToC       Page 1 
Internet Engineering Task Force (IETF)                          Z. Zhang
Request for Comments: 7716                                   L. Giuliano
Category: Standards Track                                  E. Rosen, Ed.
ISSN: 2070-1721                                   Juniper Networks, Inc.
                                                          K. Subramanian
                                                     Cisco Systems, Inc.
                                                              D. Pacella
                                                                 Verizon
                                                           December 2015


  Global Table Multicast with BGP Multicast VPN (BGP-MVPN) Procedures

Abstract

   RFCs 6513, 6514, and others describe protocols and procedures that a
   Service Provider (SP) may deploy in order to offer Multicast Virtual
   Private Network (Multicast VPN or MVPN) service to its customers.
   Some of these procedures use BGP to distribute VPN-specific multicast
   routing information across a backbone network.  With a small number
   of relatively minor modifications, the same BGP procedures can also
   be used to distribute multicast routing information that is not
   specific to any VPN.  Multicast that is outside the context of a VPN
   is known as "Global Table Multicast", or sometimes simply as
   "Internet multicast".  In this document, we describe the
   modifications that are needed to use the BGP-MVPN procedures for
   Global Table Multicast.

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/rfc7716.

Page 2 
Copyright Notice

   Copyright (c) 2015 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  . . . . . . . . . . . . . . . . . . . . . . . .   3
   2.  Adapting MVPN Procedures to GTM . . . . . . . . . . . . . . .   5
     2.1.  Use of Route Distinguishers . . . . . . . . . . . . . . .   5
     2.2.  Use of Route Targets  . . . . . . . . . . . . . . . . . .   6
     2.3.  UMH-Eligible Routes . . . . . . . . . . . . . . . . . . .   8
       2.3.1.  Routes of SAFI 1, 2, or 4 with MVPN ECs . . . . . . .   9
       2.3.2.  MVPN ECs on the Route to the Next Hop . . . . . . . .   9
       2.3.3.  Non-BGP Routes as the UMH-Eligible Routes . . . . . .  11
       2.3.4.  Why SFS Does Not Apply to GTM . . . . . . . . . . . .  11
     2.4.  Inclusive and Selective Tunnels . . . . . . . . . . . . .  13
     2.5.  I-PMSI A-D Routes . . . . . . . . . . . . . . . . . . . .  13
       2.5.1.  Intra-AS I-PMSI A-D Routes  . . . . . . . . . . . . .  13
       2.5.2.  Inter-AS I-PMSI A-D Routes  . . . . . . . . . . . . .  13
     2.6.  S-PMSI A-D Routes . . . . . . . . . . . . . . . . . . . .  14
     2.7.  Leaf A-D Routes . . . . . . . . . . . . . . . . . . . . .  14
     2.8.  Source Active A-D Routes  . . . . . . . . . . . . . . . .  14
       2.8.1.  Finding the Originator of an SA A-D Route . . . . . .  14
       2.8.2.  Optional Additional Constraints on Distribution . . .  15
     2.9.  C-Multicast Source/Shared Tree Joins  . . . . . . . . . .  16
   3.  Differences from Other MVPN-Like GTM Procedures . . . . . . .  16
   4.  Security Considerations . . . . . . . . . . . . . . . . . . .  18
   5.  References  . . . . . . . . . . . . . . . . . . . . . . . . .  19
     5.1.  Normative References  . . . . . . . . . . . . . . . . . .  19
     5.2.  Informative References  . . . . . . . . . . . . . . . . .  19
   Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . .  20
   Contributors  . . . . . . . . . . . . . . . . . . . . . . . . . .  21
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  22

Top      ToC       Page 3 
1.  Introduction

   [RFC4364] specifies architecture, protocols, and procedures that a
   Service Provider (SP) can use to provide Virtual Private Network
   (VPN) service to its customers.  In that architecture, one or more
   Customer Edge (CE) routers attach to a Provider Edge (PE) router.
   Each CE router belongs to a single VPN, but CE routers from several
   VPNs may attach to the same PE router.  In addition, CEs from the
   same VPN may attach to different PEs.  BGP is used to carry VPN-
   specific information among the PEs.  Each PE router maintains a
   separate Virtual Routing and Forwarding table (VRF) for each VPN to
   which it is attached.

   [RFC6513] and [RFC6514] extend the procedures of [RFC4364] to allow
   the SP to provide multicast service to its VPN customers.  The
   customer's multicast routing protocol (e.g., PIM) is used to exchange
   multicast routing information between a CE and a PE.  The PE stores a
   given customer's multicast routing information in the VRF for that
   customer's VPN.  BGP is used to distribute certain multicast-related
   control information among the PEs that attach to a given VPN, and BGP
   may also be used to exchange the customer multicast routing
   information itself among the PEs.

   While this multicast architecture was originally developed for VPNs,
   it can also be used (with a small number of modifications to the
   procedures) to distribute multicast routing information that is not
   specific to VPNs.  The purpose of this document is to specify the way
   in which BGP-MVPN procedures can be adapted to support non-VPN
   multicast.

   Multicast routing information that is not specific to VPNs is stored
   in a router's "global table", rather than in a VRF; hence, it is
   known as "Global Table Multicast" (GTM).  GTM is sometimes more
   simply called "Internet multicast".  However, we will avoid that term
   because it suggests that the multicast data streams are available on
   the "public" Internet.  The procedures for GTM can certainly be used
   to support multicast on the public Internet, but they can also be
   used to support multicast streams that are not public, e.g., content
   distribution streams offered by content providers to paid
   subscribers.  For the purposes of this document, all that matters is
   that the multicast routing information is maintained in a global
   table rather than in a VRF.

   This architecture does assume that the network over which the
   multicast streams travel can be divided into a "core network" and one
   or more non-core parts of the network, which we shall call
   "attachment networks".  The multicast routing protocol used in the
   attachment networks may not be the same as the one used in the core,

Top      ToC       Page 4 
   so we consider there to be a "protocol boundary" between the core
   network and the attachment networks.  We will use the term "Protocol
   Boundary Router" (PBR) to refer to the core routers that are at the
   boundary.  We will use the term "Attachment Router" (AR) to refer to
   the routers that attach to the PBRs but are not in the core.

   This document does not make any particular set of assumptions about
   the protocols that the ARs and the PBRs use to exchange unicast and
   multicast routing information with each other.  For instance,
   multicast routing information could be exchanged between an AR and a
   PBR via PIM, IGMP, or even BGP.  Multicast routing also depends on an
   exchange of routes that are used for looking up the path to the root
   of a multicast tree.  This routing information could be exchanged
   between an AR and a PBR via IGP, via EBGP, or via IBGP [RFC6368].
   Note that if IBGP is used, the "push" and "pop" procedures described
   in [RFC6368] are not necessary.

   The PBRs are not necessarily "edge routers", in the sense of
   [RFC4364].  For example, they may be both be Autonomous System Border
   Routers (ASBRs).  As another example, an AR may be an "access router"
   attached to a PBR that is an OSPF Area Border Router (ABR).  Many
   other deployment scenarios are possible.  However, the PBRs are
   always considered to be delimiting a "backbone" or "core" network.  A
   multicast data stream from an AR is tunneled over the core network
   from an ingress PBR to one or more egress PBRs.  Multicast routing
   information that a PBR learns from the ARs attached to it is stored
   in the PBR's global table.  The PBRs use BGP to distribute multicast
   routing and auto-discovery information among themselves.  This is
   done following the procedures of [RFC6513], [RFC6514], and other MVPN
   specifications, as modified in this document.

   In general, PBRs follow the same BGP-MVPN procedures that PE routers
   follow, except that these procedures are adapted to be applicable to
   the global table rather than to a VRF.  Details are provided in
   subsequent sections of this document.

   By supporting GTM using the BGP procedures designed for MVPN, one
   obtains a single control plane that governs the use of both VPN and
   non-VPN multicast.  Most of the features and characteristics of MVPN
   carry over automatically to GTM.  These include scaling, aggregation,
   flexible choice of tunnel technology in the SP network, support for
   both segmented and non-segmented tunnels, ability to use wildcards to
   identify sets of multicast flows, support for the Any-Source
   Multicast (ASM), Source-Specific Multicast (SSM), and Bidirectional
   (bidir) multicast paradigms, support for both IPv4 and IPv6 multicast
   flows over either an IPv4 or IPv6 SP infrastructure, support for
   unsolicited flooded data (including support for Bootstrap Router
   (BSR) as an RP-to-group mapping protocol), etc.

Top      ToC       Page 5 
   This document not only uses MVPN procedures for GTM but also, insofar
   as possible, uses the same protocol elements, encodings, and formats.
   The BGP Updates for GTM thus use the same Subsequent Address Family
   Identifier (SAFI) and have the same Network Layer Reachability
   Information (NLRI) format as the BGP Updates for MVPN.

   Details for supporting MVPN (either IPv4 or IPv6 MVPN traffic) over
   an IPv6 backbone network can be found in [RFC6515].  The procedures
   and encodings described therein are also applicable to GTM.

   [RFC7524] extends [RFC6514] by providing procedures that allow
   tunnels through the core to be "segmented" at ABRs within the core.
   The ABR segmentation procedures are also applicable to GTM as defined
   in the current document.  In general, the MVPN procedures of
   [RFC7524], adapted as specified in the current document, are
   applicable to GTM.

   [RFC7524] also defines a set of procedures for GTM.  Those procedures
   are different from the procedures defined in the current document,
   and the two sets of procedures are not interoperable with each other.
   The two sets of procedures can co-exist in the same network, as long
   as they are not applied to the same multicast flows or to the same
   multicast group addresses.  See Section 3 for more details.

   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.  Adapting MVPN Procedures to GTM

   In general, PBRs support Global Table Multicast by using the
   procedures that PE routers use to support VPN multicast.  For GTM,
   where [RFC6513] and [RFC6514] talk about the "PE-CE interface", one
   should interpret that to mean the interface between the AR and the
   PBR.  For GTM, where [RFC6513] and [RFC6514] talk about the
   "backbone" network, one should interpret that to mean the part of the
   network that is delimited by the PBRs.

   A few adaptations to the procedures of [RFC6513] and [RFC6514] need
   to be made.  Those adaptations are described in the following sub-
   sections.

2.1.  Use of Route Distinguishers

   The MVPN procedures require the use of BGP routes, defined in
   [RFC6514], that have a SAFI value of 5 ("MCAST-VPN").  We refer to
   these simply as "MCAST-VPN routes".  [RFC6514] defines the Network
   Layer Reachability Information (NLRI) format for MCAST-VPN routes.

Top      ToC       Page 6 
   The NLRI field always begins with a "Route Type" octet, and,
   depending on the route type, may be followed by a Route Distinguisher
   (RD) field.

   When a PBR originates an MCAST-VPN route in support of GTM, the RD
   field (for those routes types where it is defined) of that route's
   NLRI MUST be set to zero (i.e., to 64 bits of zero).  Since no VRF
   may have an RD of zero, this allows MCAST-VPN routes that are about
   GTM to be distinguished from MCAST-VPN routes that are about VPNs.

2.2.  Use of Route Targets

   The MVPN procedures require all MCAST-VPN routes to carry Route
   Targets (RTs).  When a PE router receives an MCAST-VPN route, it
   processes the route in the context of a particular VRF if and only if
   the route is carrying an RT that is configured as one of that VRF's
   "import RTs".

   There are two different kinds of RT used in MVPN.

   o  One kind of RT is carried only by the following MCAST-VPN route
      types: C-multicast Shared Tree Joins, C-multicast Source Tree
      Joins, and Leaf auto-discovery routes (A-D routes).  This kind of
      RT identifies the PE router that has been selected by the route's
      originator as the "Upstream PE" or as the "Upstream Multicast Hop"
      (UMH) for a particular (set of) multicast flow(s).  Per [RFC6514]
      and [RFC6515], this RT must be an IPv4-address-specific or IPv6-
      address-specific Extended Community (EC), whose Global
      Administrator field identifies the Upstream PE or the UMH.  If the
      Global Administrator field identifies the Upstream PE, the Local
      Administrator field identifies a particular VRF in that PE.

      The GTM procedures of this document require the use of this type
      of RT, in exactly the same situations where it is used in the MVPN
      specification [RFC6514].  However, one adaptation is necessary:
      the Local Administrator field of this kind of RT MUST always be
      set to zero, thus implicitly identifying the global table rather
      than identifying a VRF.  We will refer to this kind of RT as an
      "upstream-node-identifying RT".

   o  The other kind of RT is the conventional RT first specified in
      [RFC4364].  It does not necessarily identify a particular router
      by address but is used to constrain the distribution of VPN routes
      and to ensure that a given VPN route is processed in the context
      of a given VRF if and only if the route is carrying an RT that has
      been configured as one of that VRF's "import RTs".

Top      ToC       Page 7 
      Whereas every VRF must be configured with at least one import RT,
      there has been no requirement to configure any RTs for the global
      table of any router until now.  As stated above, this document
      makes the use of upstream-node-identifying RTs mandatory for GTM.
      This document makes the use of non-upstream-node-identifying RTs
      OPTIONAL for GTM.

   The procedures for the use of RTs in GTM are the following:

   o  If the global table of a particular PBR is NOT configured with any
      import RTs, then a received MCAST-VPN route is processed in the
      context of the global table only if it is carrying no RTs or if it
      is carrying an upstream-node-identifying RT whose Global
      Administrator field identifies that PBR.

   o  The global table in each PBR MAY be configured with (a) a set of
      export RTs to be attached to MCAST-VPN routes that are originated
      to support GTM and (b) a set of import RTs for GTM.

      If the global table of a given PBR has been so configured, the PBR
      will process a received MCAST-VPN route in the context of the
      global table if and only if the route carries an RT that is one of
      the global table's import RTs or if the route carries an upstream-
      node-identifying RT whose Global Administrator field identifies
      the PBR.

      If the global tables are configured with RTs, care must be taken
      to ensure that the RTs configured for the global table are
      distinct from any RTs used in support of MVPN (except in the case
      where it is actually intended to create an "extranet"
      [MVPN-extranet] in which some sources are reachable in global
      table context while others are reachable in VPN context.)

   The "RT Constraint" procedures of [RFC4684] MAY be used to constrain
   the distribution of MCAST-VPN routes (or other routes) that carry RTs
   that have been configured as import RTs for GTM.  (This includes the
   upstream-node-identifying RTs.)

   N.B.: If the "RT Constraint" procedures of [RFC4684] are deployed,
         but the MCAST-VPN routes are not carrying RTs, then proper
         operation requires the "default behavior" specified for the
         MCAST-VPN address family in Section 3 ("Default Behavior") of
         [RTC_without_RTs].

   In [RFC6513], the UMH-eligible routes (see Section 5.1.1 of
   [RFC6513], "Eligible Routes for UMH Selection") are generally routes
   of SAFI 128 (Labeled VPN-IP routes) or 129 (VPN-IP multicast routes)
   and are required to carry RTs.  These RTs determine which VRFs import

Top      ToC       Page 8 
   which such routes.  However, for GTM, when the UMH-eligible routes
   may be routes of SAFI 1, 2, or 4, the routes are not required to
   carry RTs.  This document does NOT specify any new rules for
   determining whether a SAFI 1, 2, or 4 route is to be imported into
   the global table of any PBR.

2.3.  UMH-Eligible Routes

   Section 5.1 of [RFC6513] defines procedures by which a PE router
   determines the "C-root", the "Upstream Multicast Hop" (UMH), the
   "Upstream PE", and the "Upstream RD" of a given multicast flow.  (In
   non-VPN multicast documents, the UMH of a multicast flow at a
   particular router is generally known as the "RPF neighbor" for that
   flow.)  It also defines procedures for determining the "Source AS" of
   a particular flow.  Note that in GTM, the "Upstream PE" is actually
   the "Upstream PBR".

   The definition of the C-root of a flow is the same for GTM as for
   MVPN.

   For MVPN, to determine the UMH, Upstream PE, Upstream RD, and Source
   AS of a flow, one looks up the C-root of the flow in a particular VRF
   and finds the "UMH-eligible" routes (see Section 5.1.1 of [RFC6513])
   that "match" the C-root.  From among these, one is chosen as the
   "Selected UMH Route".

   For GTM, the C-root is, of course, looked up in the global table,
   rather than in a VRF.  For MVPN, the UMH-eligible routes are routes
   of SAFI 128 or 129.  For GTM, the UMH-eligible routes are routes of
   SAFI 1, SAFI 4, or SAFI 2.  If the global table has imported routes
   of SAFI 2, then these are the UMH-eligible routes.  Otherwise, routes
   of SAFI 1 or SAFI 4 are the UMH-eligible routes.  For the purpose of
   UMH determination, if a SAFI 1 route and a SAFI 4 route contain the
   same IP prefix in their respective NLRI fields, then the two routes
   are considered by the BGP best-path selection process to be
   comparable.

   [RFC6513] defines procedures for determining which of the UMH-
   eligible routes that match a particular C-root is to become the
   Selected UMH Route.  With one exception, these procedures are also
   applicable to GTM.  The one exception is the following.
   Section 9.1.2 of [RFC6513] defines a particular method of choosing
   the Upstream PE, known as "Single Forwarder Selection" (SFS).  This
   procedure MUST NOT be used for GTM (see Section 2.3.4 for an
   explanation of why the SFS procedure cannot be applied to GTM).

   In GTM, the "Upstream RD" of a multicast flow is always considered to
   be zero and is NOT determined from the Selected UMH Route.

Top      ToC       Page 9 
   The MVPN specifications require that when BGP is used for
   distributing multicast routing information, the UMH-eligible routes
   MUST carry the VRF Route Import EC and the Source AS EC.  To
   determine the Upstream PE and Source AS for a particular multicast
   flow, the Upstream PE and Source AS are determined, respectively,
   from the VRF Route Import EC and the Source AS EC of the Selected UMH
   Route for that flow.  These ECs are generally attached to the UMH-
   eligible routes by the PEs that originate the routes.

   In GTM, there are certain situations in which it is allowable to omit
   the VRF Route Import EC and/or the Source AS EC from the UMH-eligible
   routes.  The following sub-sections specify the various options for
   determining the Upstream PBR and the Source AS in GTM.

   The procedures in Section 2.3.1 MUST be implemented.  The procedures
   in Sections 2.3.2 and 2.3.3 are OPTIONAL to implement.  It should be
   noted that while the optional procedures may be useful in particular
   deployment scenarios, there is always the potential for
   interoperability problems when relying on OPTIONAL procedures.

2.3.1.  Routes of SAFI 1, 2, or 4 with MVPN ECs

   If the UMH-eligible routes have a SAFI of 1, 2, or 4, then they MAY
   carry the VRF Route Import EC and/or the Source AS EC.  If the
   Selected UMH Route is a route of SAFI 1, 2, or 4 that carries the VRF
   Route Import EC, then the Upstream PBR is determined from that EC.
   Similarly, if the Selected UMH Route is a route of SAFI 1, 2, or 4
   that carries the Source AS EC, the Source AS is determined from that
   EC.

   When the procedure of this section is used, a PBR that distributes a
   UMH-eligible route to other PBRs is responsible for ensuring that the
   VRF Route Import and Source AS ECs are attached to it.

   If the selected UMH-eligible route has a SAFI of 1, 2, or 4 but is
   not carrying a VRF Route Import EC, then the Upstream PBR is
   determined as specified in Sections 2.3.2 or 2.3.3.

   If the selected UMH-eligible route has a SAFI of 1, 2, or 4 but is
   not carrying a Source AS EC, then the Source AS is considered to be
   the local AS.

2.3.2.  MVPN ECs on the Route to the Next Hop

   Some service providers may consider it to be undesirable to have the
   PBRs put the VRF Route Import EC on all the UMH-eligible routes.  Or
   there may be deployment scenarios in which the UMH-eligible routes
   are not advertised by the PBRs at all.  The procedures described in

Top      ToC       Page 10 
   this section provide an alternative that can be used under certain
   circumstances.

   The procedures in this section are OPTIONAL.

   In this alternative procedure, each PBR MUST originate a BGP route of
   SAFI 1, 2, or 4 whose NLRI is an IP address of the PBR itself.  This
   route MUST carry a VRF Route Import EC that identifies the PBR.  The
   address that appears in the Global Administrator field of that EC
   MUST be the same address that appears in the NLRI and in the Next Hop
   field of that route.  This route MUST also carry a Source AS EC
   identifying the AS of the PBR.

   Whenever the PBR distributes a UMH-eligible route for which it sets
   itself as Next Hop, it MUST use this same IP address as the Next Hop
   of the UMH-eligible route that it used in the route discussed in the
   prior paragraph.

   When the procedure in this section is used and when a PBR is
   determining the Selected UMH Route for a given multicast flow, it may
   find that the Selected UMH Route has no VRF Route Import EC.  In this
   case, the PBR will look up (in the global table) the route to the
   Next Hop of the Selected UMH Route.  If the route to the Next Hop has
   a VRF Route Import EC, that EC will be used to determine the Upstream
   PBR, just as if the EC had been attached to the Selected UMH Route.

   If recursive route resolution is required in order to resolve the
   Next Hop, the Upstream PBR will be determined from the first route
   with a VRF Route Import EC that is encountered during the recursive
   route resolution process.  (The recursive route resolution process
   itself is not modified by this document.)

   The same procedure can be applied to find the Source AS, except that
   the Source AS EC is used instead of the VRF Route Import EC.

   Note that this procedure is only applicable in scenarios where it is
   known that the Next Hop of the UMH-eligible routes is not changed by
   any router that participates in the distribution of those routes;
   this procedure MUST NOT be used in any scenario where the Next Hop
   may be changed between the time one PBR distributes the route and
   another PBR receives it.  The PBRs have no way of determining
   dynamically whether the procedure is applicable in a particular
   deployment; this must be made known to the PBRs by provisioning.

   Some scenarios in which this procedure can be used are:

   o  All PBRs are in the same AS.

Top      ToC       Page 11 
   o  The UMH-eligible routes are distributed among the PBRs by a Route
      Reflector (that does not change the Next Hop).

   o  The UMH-eligible routes are distributed from one AS to another
      through ASBRs that do not change the Next Hop.

   If the procedures of this section are used in scenarios where they
   are not applicable, GTM will not function correctly.

2.3.3.  Non-BGP Routes as the UMH-Eligible Routes

   In particular deployment scenarios, there may be specific procedures
   that can be used, in those particular scenarios, to determine the
   Upstream PBR for a given multicast flow.

   Suppose the PBRs neither put the VRF Route Import EC on the UMH-
   eligible routes, nor distribute BGP routes with their own addresses
   in the NLRI.  It may still be possible, by using specific knowledge
   about the deployment, to determine the Upstream PBR for a given
   multicast flow.

   For example, suppose it is known that all the PBRs are in the same
   OSPF area.  It may be possible to determine the Upstream PBR for a
   given multicast flow by looking at the link state database to see
   which router is attached to the flow's C-root.

   As another example, suppose it is known that the set of PBRs is fully
   meshed via Traffic Engineering (TE) tunnels.  When a PBR looks up, in
   its global table, the C-root of a particular multicast flow, it may
   find that the next-hop interface is a particular TE tunnel.  If it
   can determine the identity of the router at the other end of that TE
   tunnel, it can deduce that the router is the Upstream PBR for that
   flow.

   This is not an exhaustive set of examples.  Any procedure that
   correctly determines the Upstream PBR in a given deployment scenario
   MAY be used in that scenario.

2.3.4.  Why SFS Does Not Apply to GTM

   To see why the SFS procedure cannot be applied to GTM, consider the
   following example scenario.  Suppose some multicast source S is homed
   to both PBR1 and PBR2, and suppose that both PBRs export a route (of
   SAFI 1, 2, or 4) whose NLRI is a prefix matching the address of S.
   These two routes will be considered comparable by the BGP decision
   process.  A route reflector receiving both routes may thus choose to
   redistribute just one of the routes to S, the one chosen by the best-
   path algorithm.  Different route reflectors may even choose different

Top      ToC       Page 12 
   routes to redistribute (i.e., one route reflector may choose the
   route to S via PBR1 as the best path, while another chooses the route
   to S via PBR2 as the best path).  As a result, some PBRs may receive
   only the route to S via PBR1, and some may receive only the route to
   S via PBR2.  In that case, it is impossible to ensure that all PBRs
   will choose the same route to S.

   The SFS procedure works in VPN context as long as the following
   assumption holds: if S is homed to VRF-x in PE1 and to VRF-y in PE2,
   then VRF-x and VRF-y have been configured with different RDs.  In VPN
   context, the route to S is of SAFI 128 or 129 and thus has an RD in
   its NLRI.  So the route to S via PE1 will not have the same NLRI as
   the route to S via PE2.  As a result, all PEs will see both routes,
   and the PEs can implement a procedure that ensures that they all pick
   the same route to S.

   That is, the SFS procedure of [RFC6513] relies on the UMH-eligible
   routes being of SAFI 128 or 129 and relies on certain VRFs being
   configured with distinct RDs.  Thus, the procedure cannot be applied
   to GTM.

   One might think that the SFS procedure could be applied to GTM as
   long as the procedures defined in [ADD-PATH] are applied to the UMH-
   eligible routes.  Using the [ADD-PATH] procedures, the BGP speakers
   could advertise more than one path to a given prefix.  Typically,
   [ADD-PATH] is used to report the n best paths, for some small value
   of n.  However, this is not sufficient to support SFS, as can be seen
   by examining the following scenario:

                              AS-X  |   AS-Y  |    AS-Z
                                    |         |
                   S--PBR1---ASBR1--|--ASBR2--|---ASBR5
                   |   \______/     |         |
                   |   /      \     |         |
                   |--PBR2---ASBR3--|--ASBR4--|---ASBR6
                                    |         |

   In AS-X, PBR1 reports to both ASBR1 and ASBR3 that it has a route to
   S.  Similarly, PBR2 reports to both ASBR1 and ASBR3 that it has a
   route to S.  Using the procedures in [ADD-PATH], ASBR1 reports both
   routes to ASBR2, and ASBR3 reports both routes to ASBR4.  Now AS-Y
   sees 4 paths to S.  The AS-Z ASBRs will each see eight paths (four
   via ASBR2 and four via ASBR4).  To avoid this explosion in the number
   of paths, a BGP speaker that uses [ADD-PATH] is usually considered to
   report only the n best paths.  However, there is then no guarantee
   that the reported set of paths will contain at least one path via
   PBR1 and at least one path via PBR2.  Without such a guarantee, the
   SFS procedure will not work.

Top      ToC       Page 13 
2.4.  Inclusive and Selective Tunnels

   The MVPN specifications allow multicast flows to be carried on either
   Inclusive Tunnels or on Selective Tunnels.  When a flow is sent on an
   Inclusive Tunnel of a particular VPN, it is sent to all PEs in that
   VPN.  When sent on a Selective Tunnel of a particular VPN, it may be
   sent to only a subset of the PEs in that VPN.

   This document allows the use of either Inclusive Tunnels or Selective
   Tunnels for GTM.  However, any service provider electing to use
   Inclusive Tunnels for GTM should carefully consider whether sending a
   multicast flow to ALL its PBRs would result in problems of scale.
   There are potentially many more PBRs for GTM than PEs for a
   particular VPN.  If the set of PBRs is large and growing, but most
   multicast flows do not need to go to all the PBRs, the exclusive use
   of Selective Tunnels may be a better option.

2.5.  I-PMSI A-D Routes

2.5.1.  Intra-AS I-PMSI A-D Routes

   Per [RFC6514], there are certain conditions under which it is NOT
   required for a PE router implementing MVPN to originate one or more
   Intra-AS Inclusive Provider Multicast Service Interface (I-PMSI) A-D
   routes.  These conditions also apply to PBRs implementing GTM.

   In addition, a PBR implementing GTM is NOT required to originate an
   Intra-AS I-PMSI A-D route if both of the following conditions hold:

   o  The PBR is not using Inclusive Tunnels for GTM, and

   o  The distribution of the C-multicast Shared Tree Join and
      C-multicast Source Tree Join routes is done in such a manner that
      the Next Hop of those routes does not change.

   Please see also the sections on RD and RT usage (Sections 2.1 and
   2.2, respectively).

2.5.2.  Inter-AS I-PMSI A-D Routes

   There are no GTM-specific procedures for the origination,
   distribution, and processing of these routes, other than those
   specified in the sections on RD and RT usage (Sections 2.1 and 2.2).

Top      ToC       Page 14 
2.6.  S-PMSI A-D Routes

   There are no GTM-specific procedures for the origination,
   distribution, and processing of these routes, other than those
   specified in the sections on RD and RT usage (Sections 2.1 and 2.2).

2.7.  Leaf A-D Routes

   There are no GTM-specific procedures for the origination,
   distribution, and processing of these routes, other than those
   specified in the sections on RD and RT usage (Sections 2.1 and 2.2).

2.8.  Source Active A-D Routes

   Please see the sections on RD and RT usage (Sections 2.1 and 2.2) for
   information that applies to the origination and distribution of
   Source Active A-D routes.  Additional procedures governing the use of
   Source Active A-D routes are given in the sub-sections of this
   section.

2.8.1.  Finding the Originator of an SA A-D Route

   To carry out the procedures specified in [RFC6514] (e.g., in
   Section 13.2 of that document), it is sometimes necessary for an
   egress PE to determine the ingress PE that originated a given Source
   Active A-D route.  The procedure used in [RFC6514] to find the
   originator of a Source Active A-D route assumes that no two routes
   have the same RD unless they have been originated by the same PE.
   However, this assumption is not valid in GTM, because each Source
   Active A-D route used for GTM will have an RD of 0, and all the UMH-
   eligible routes also have an RD of 0.  So GTM requires a different
   procedure for determining the originator of a Source Active A-D
   route.

   In GTM, the procedure for determining the originating PE of a Source
   Active A-D route is the following:

   o  When a Source Active A-D route is originated, the originating PE
      MAY attach a VRF Route Import Extended Community to the route.

   o  When a Source Active A-D route is distributed by one BGP speaker
      to another, then:

      *  If the Source Active A-D route does not carry the VRF Route
         Import EC, the BGP speaker distributing the route MUST NOT
         change the route's Next Hop field.

Top      ToC       Page 15 
      *  If the Source Active A-D route does carry the VRF Route Import
         EC, the BGP speaker distributing the route MAY change the
         route's Next Hop field to itself.

   o  When an egress PE needs to determine the originator of a Source
      Active A-D route, then:

      *  If the Source Active A-D route carries the VRF Route Import EC,
         the originating PE is the PE identified in the Global
         Administrator field of that EC.

      *  If the Source Active A-D route does not carry the VRF Route
         Import EC, the originating PE is the PE identified in the
         route's Next Hop field.

2.8.2.  Optional Additional Constraints on Distribution

   If some site has receivers for a particular ASM group G, then it is
   possible (by the procedures of [RFC6514]) that every PBR attached to
   a site with a source for group G will originate a Source Active A-D
   route whose NLRI identifies that source and group.  These Source
   Active A-D routes may be distributed to every PBR.  If only a
   relatively small number of PBRs are actually interested in traffic
   from group G, but there are many sources for group G, this could
   result in a large number of (S,G) Source Active A-D routes being
   installed in a large number of PBRs that have no need of them.

   For GTM, it is possible to constrain the distribution of (S,G) Source
   Active A-D routes to those PBRs that are interested in GTM traffic to
   group G.  This can be done using the following OPTIONAL procedures:

   o  If a PBR originates a C-multicast Shared Tree Join whose NLRI
      contains (RD=0,*,G), then it dynamically creates an import RT for
      its global table, where the Global Administrator field of the RT
      contains the group address G, and the Local Administrator field
      contains zero.  (Note that an IPv6-address-specific RT would need
      to be used if the group address is an IPv6 address.)

   o  When a PBR creates such an import RT, it uses "RT Constraint"
      procedures [RFC4684] to advertise its interest in routes that
      carry this RT.

   o  When a PBR originates a Source Active A-D route from its global
      table, it attaches the RT described above.

   o  When the C-multicast Shared Tree Join is withdrawn, so is the
      corresponding RT constrain route, and the corresponding RT is
      removed as an import RT of its global table.

Top      ToC       Page 16 
   These procedures enable a PBR to automatically filter all Source
   Active A-D routes that are about multicast groups in which the PBR
   has no interest.

   This procedure does introduce the overhead of distributing additional
   "RT Constraint" routes and therefore may not be cost-effective in all
   scenarios, especially if the number of sources per ASM group is
   small.  This procedure may also result in increased join latency.

2.9.  C-Multicast Source/Shared Tree Joins

   Section 11.1.3 of [RFC6514] describes how to determine the IP-
   address-specific RT(s) that should be attached to a C-multicast
   route.  The "Upstream PE", "Upstream RD", and "Source AS" (as defined
   in Section 5 of [RFC6513]) for the NLRI of the C-multicast route are
   first determined.  An IP-address-specific RT whose Global
   Administrator field carries the IP address of the Upstream PE is then
   attached to the C-multicast route.  This procedure also applies to
   GTM, except that the "Upstream PE" is actually an "Upstream PBR".

   Section 11.1.3 of [RFC6514] also specifies that a second IP-address-
   specific RT be attached to the C-multicast route, if the Source AS of
   the NLRI of that route is different than the AS of the PE originating
   the route.  The procedure for creating this RT may be summarized as:

   (a)  Determine the Upstream PE, Upstream RD, and Source AS
        corresponding to the NLRI of the route.

   (b)  Find the corresponding Inter-AS or Intra-AS I-PMSI A-D route
        based on (a).

   (c)  Find the Next Hop of that A-D route.

   (d)  Place the IP address of that Next Hop in the Global
        Administrator field of the RT.

   However, for GTM, in scenarios where it is known a priori by a PBR
   that the Next Hop of the C-multicast Source/Shared Tree Joins does
   not change during the distribution of those routes, the second RT
   (the one based on the Next Hop of an I-PMSI A-D route) is not needed
   and should not be present.  In other scenarios, the procedure of
   Section 11.1.3 of [RFC6514] (as modified by Sections 2.1 and 2.2 of
   this document) is applied by the PBRs.

3.  Differences from Other MVPN-Like GTM Procedures

   [RFC7524] also defines a procedure for GTM that is based on the BGP
   procedures that were developed for MVPN.

Top      ToC       Page 17 
   However, the GTM procedures of [RFC7524] are different than and are
   NOT interoperable with the procedures defined in this document.

   The two sets of procedures can co-exist in the same network, as long
   as they are not applied to the same multicast flows or to the same
   ASM multicast group addresses.

   Some of the major differences between the two sets of procedures are
   the following:

   o  The procedures for GTM described in [RFC7524] do not use
      C-multicast Shared Tree Joins or C-multicast Source Tree Joins at
      all.  The procedures of this document use these C-multicast routes
      for GTM, setting the RD field of the NLRI to zero.

   o  The procedures for GTM described in [RFC7524] use Leaf A-D routes
      instead of C-multicast Shared/Source Tree Join routes.  Leaf A-D
      routes used in that manner can be distinguished from Leaf A-D
      routes used as specified in [RFC6514] by means of the NLRI format;
      [RFC7524] defines a new NLRI format for Leaf A-D routes.  Whether
      or not a given Leaf A-D route is being used according to the
      procedures described in [RFC7524] can be determined from its NLRI.
      (See Section 6.2.2 ("Leaf A-D Route for Global Table Multicast")
      of [RFC7524]).

   o  The Leaf A-D routes used by the current document contain an NLRI
      that is in the format defined in [RFC6514], NOT in the format as
      defined in [RFC7524].  The procedures assumed by this document for
      originating and processing Leaf A-D routes are as specified in
      [RFC6514], NOT as specified in [RFC7524].

   o  The current document uses an RD value of zero in the NLRI in order
      to indicate that a particular route is about a Global
      Table Multicast rather than a VPN multicast.  No other semantics
      are inferred from the fact that RD is zero.  [RFC7524] uses two
      different RD values in its GTM procedures, with semantic
      differences that depend upon the RD values.

   o  In order for both sets of procedures to co-exist in the same
      network, the PBRs MUST be provisioned so that for any given IP
      group address in the global table, all egress PBRs use the same
      set of procedures for that group address (i.e., for group G,
      either all egress PBRs use the GTM procedures of this document or
      all egress PBRs use the GTM procedures of [RFC7524]).

Top      ToC       Page 18 
4.  Security Considerations

   The security considerations of this document are primarily the
   security considerations of the base protocols, as discussed in
   [RFC6514], [RFC4601], and [RFC5294].

   The protocols and procedures described in this document make use of a
   type of route (routes with the "MCAST-VPN" BGP SAFI) that was
   originally designed for use in VPN contexts only.  The protocols and
   procedures described in this document also make use of various BGP
   path attributes and extended communities (VRF Route Import Extended
   Community, Source AS Extended Community, and Route Target Extended
   Community) that were originally intended for use in VPN contexts.  If
   these routes, attributes, and/or extended communities leak out into
   the wild, multicast data flows may be distributed in an unintended
   and/or unauthorized manner.

   When VPNs are provisioned, each VRF is configured with import RTs and
   export RTs.  These RTs constrain the distribution and the import of
   the VPN routes, making it difficult to cause a route to be
   distributed to and imported by a VRF that is not authorized to import
   that route.  Additionally, VPN routes do not go into the "global
   table" with the "ordinary Internet routes" (i.e., non-VPN routes),
   and non-VPN routes do not impact the flow of multicast data within a
   VPN.  In GTM, some of these protections against improper
   distribution/import of the routes is lost -- import of the routes is
   not restricted to VRFs, and the RT infrastructure that controls the
   distribution of routes among the VRFs is not present when routes are
   exported from and imported into global tables.

   Internet Service Providers often make extensive use of BGP extended
   communities, sometimes adding, deleting, and/or modifying a route's
   extended communities as the route is distributed throughout the
   network.  Care should be taken to avoid deleting or modifying the VRF
   Route Import Extended Community and Source AS Extended Community.
   Incorrect manipulation of these extended communities may result in
   multicast streams being lost or misrouted.

   The procedures of this document require certain BGP routes to carry
   IP multicast group addresses.  Generally, such group addresses are
   only valid within a certain scope.  If a BGP route containing a group
   address is distributed outside the boundaries where the group address
   is meaningful, unauthorized distribution of multicast data flows may
   occur.

Top      ToC       Page 19 
5.  References

5.1.  Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <http://www.rfc-editor.org/info/rfc2119>.

   [RFC4364]  Rosen, E. and Y. Rekhter, "BGP/MPLS IP Virtual Private
              Networks (VPNs)", RFC 4364, DOI 10.17487/RFC4364, February
              2006, <http://www.rfc-editor.org/info/rfc4364>.

   [RFC6513]  Rosen, E., Ed. and R. Aggarwal, Ed., "Multicast in MPLS/
              BGP IP VPNs", RFC 6513, DOI 10.17487/RFC6513, February
              2012, <http://www.rfc-editor.org/info/rfc6513>.

   [RFC6514]  Aggarwal, R., Rosen, E., Morin, T., and Y. Rekhter, "BGP
              Encodings and Procedures for Multicast in MPLS/BGP IP
              VPNs", RFC 6514, DOI 10.17487/RFC6514, February 2012,
              <http://www.rfc-editor.org/info/rfc6514>.

   [RFC6515]  Aggarwal, R. and E. Rosen, "IPv4 and IPv6 Infrastructure
              Addresses in BGP Updates for Multicast VPN", RFC 6515,
              DOI 10.17487/RFC6515, February 2012,
              <http://www.rfc-editor.org/info/rfc6515>.

5.2.  Informative References

   [ADD-PATH] Walton, D., Retana, A., Chen, E., and J. Scudder,
              "Advertisement of Multiple Paths in BGP", Work in
              Progress, draft-ietf-idr-add-paths-12, November 2015.

   [MVPN-extranet]
              Rekhter, Y., Rosen, E., Aggarwal, R., Cai, Y., and T.
              Morin, "Extranet Multicast in BGP/IP MPLS VPNs", Work in
              Progress, draft-ietf-bess-mvpn-extranet-04, November 2015.

   [RFC4601]  Fenner, B., Handley, M., Holbrook, H., and I. Kouvelas,
              "Protocol Independent Multicast - Sparse Mode (PIM-SM):
              Protocol Specification (Revised)", RFC 4601,
              DOI 10.17487/RFC4601, August 2006,
              <http://www.rfc-editor.org/info/rfc4601>.

Top      ToC       Page 20 
   [RFC4684]  Marques, P., Bonica, R., Fang, L., Martini, L., Raszuk,
              R., Patel, K., and J. Guichard, "Constrained Route
              Distribution for Border Gateway Protocol/MultiProtocol
              Label Switching (BGP/MPLS) Internet Protocol (IP) Virtual
              Private Networks (VPNs)", RFC 4684, DOI 10.17487/RFC4684,
              November 2006, <http://www.rfc-editor.org/info/rfc4684>.

   [RFC5294]  Savola, P. and J. Lingard, "Host Threats to Protocol
              Independent Multicast (PIM)", RFC 5294,
              DOI 10.17487/RFC5294, August 2008,
              <http://www.rfc-editor.org/info/rfc5294>.

   [RFC6368]  Marques, P., Raszuk, R., Patel, K., Kumaki, K., and T.
              Yamagata, "Internal BGP as the Provider/Customer Edge
              Protocol for BGP/MPLS IP Virtual Private Networks (VPNs)",
              RFC 6368, DOI 10.17487/RFC6368, September 2011,
              <http://www.rfc-editor.org/info/rfc6368>.

   [RFC7524]  Rekhter, Y., Rosen, E., Aggarwal, R., Morin, T.,
              Grosclaude, I., Leymann, N., and S. Saad, "Inter-Area
              Point-to-Multipoint (P2MP) Segmented Label Switched Paths
              (LSPs)", RFC 7524, DOI 10.17487/RFC7524, May 2015,
              <http://www.rfc-editor.org/info/rfc7524>.

   [RTC_without_RTs]
              Rosen, E., Ed., Patel, K., Haas, J., and R. Raszuk, "Route
              Target Constrained Distribution of Routes with no Route
              Targets", Work in Progress, draft-ietf-idr-rtc-no-rt-04,
              November 2015.

Acknowledgments

   The authors and contributors would like to thank Rahul Aggarwal,
   Huajin Jeng, Hui Ni, Yakov Rekhter, and Samir Saad for their ideas
   and comments.

Top      ToC       Page 21 
Contributors

   Jason Schiller
   Google
   Suite 400
   1818 Library Street
   Reston, Virginia 20190
   United States

   Email: jschiller@google.com

   Zhenbin Li
   Huawei Technologies
   Huawei Blvd., No.156 Beiqing Rd.
   Beijing  100095
   China

   Email: lizhenbin@huawei.com

   Wei Meng
   ZTE Corporation
   No.50 Software Avenue, Yuhuatai District
   Nanjing
   China

   Email: meng.wei2@zte.com.cn,vally.meng@gmail.com

   Cui Wang
   ZTE Corporation
   No.50 Software Avenue, Yuhuatai District
   Nanjing
   China

   Email: wang.cui1@zte.com.cn

   Shunwan Zhuang
   Huawei Technologies
   Huawei Blvd., No.156 Beiqing Rd.
   Beijing  100095
   China

   Email: zhuangshunwan@huawei.com

Top      ToC       Page 22 
Authors' Addresses

   Zhaohui Zhang
   Juniper Networks, Inc.
   10 Technology Park Drive
   Westford, Massachusetts  01886
   United States

   Email: zzhang@juniper.net


   Lenny Giuliano
   Juniper Networks, Inc.
   2251 Corporate Park Drive
   Herndon, Virginia  20171
   United States

   Email: lenny@juniper.net


   Eric C. Rosen (editor)
   Juniper Networks, Inc.
   10 Technology Park Drive
   Westford, Massachusetts  01886
   United States

   Email: erosen@juniper.net


   Karthik Subramanian
   Cisco Systems, Inc.
   170 Tasman Drive
   San Jose, CA  95134
   United States

   Email: kartsubr@cisco.com


   Dante J. Pacella
   Verizon
   22001 Loudoun County Parkway
   Ashburn, Virginia  95134
   United States

   Email: dante.j.pacella@verizonbusiness.com