[Docs] [txt|pdf] [Tracker] [Email] [Diff1] [Diff2] [Nits]

Versions: 00 01 02 03 04 05 06

Network Working Group                                     Hannes Gredler
Internet Draft                                          Juniper Networks
Intended status: Standards Track
Expires: November 2014                                     Yakov Rekhter
                                                        Juniper Networks

                                                              Luay Jalil
                                                                 Verizon

                                                          Sriganesh Kini
                                                                Ericsson

                                                               Xiaohu Xu
                                                                  Huawei




                                                             May 21 2014


      Supporting Source/Explicitly Routed Tunnels via Stacked LSPs


                    draft-gredler-spring-mpls-06.txt

Status of this Memo

   This Internet-Draft is submitted to IETF in full conformance with the
   provisions of BCP 78 and 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.






Gredler                                                         [Page 1]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


Copyright Notice

   Copyright (c) 2013 IETF Trust and the persons identified as the
   document authors. All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document. Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document. Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.


Abstract

   This document describes how source/explicitly routed tunnels could be
   realized using stacked Label Switched Paths (LSPs).

   This document also describes how use of IS-IS/OSPF as a label
   distribution protocol fits into the MPLS architecture.




























Gredler                                                         [Page 2]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


Table of Contents

 1          Specification of Requirements  .........................   3
 2          Terminology  ...........................................   3
 3          Introduction  ..........................................   4
 4          Constructing Explicitly Routed Tunnels by using Stacked LSPs  5
 4.1        Examples of Constructing Explicitly Routed Tunnels by Stacked LSPs  8
 4.1.1      Explicitly Routed Tunnel with Single Hops  .............   8
 4.1.2      Explicitly Routed Tunnel with Multi-Hops  ..............  11
 5          IS-IS or OSPF as Label Distribution Protocol  ..........  13
 5.1        Example of IS-IS/OSPF as Label Distribution Protocols  .  15
 6          IANA Considerations  ...................................  16
 7          Security Considerations  ...............................  16
 8          Acknowledgements  ......................................  16
 9          Normative References  ..................................  16
10          Informative References  ................................  16
11          Authors' Addresses  ....................................  17






1. Specification of Requirements

   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 [RFC2119].


2. Terminology

   We use the term "explicitly routed tunnels" as a synonym for such
   terms as "source routed tunnels" and "source-initiated routed
   tunnels".

   Note that the term "source routed tunnel", or "source-initiated
   routed tunnel" does not imply that intermediate nodes of such a
   tunnel forward packets traversing the tunnel based upon source
   addresses of these packets. In the context of "source routed tunnels"
   and "source-initiated routed tunnels" the term "source" refers to the



Gredler                                                         [Page 3]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   tunnels' ingress.

   This document assumes that a reader is familiar with the MPLS
   architecture [RFC3031] terminology.

   For a given Label Switched Path (LSP) of level m, as defined in
   section 3.15 of [RFC3031]:

     + the ingress node/router is the LSR that pushes the level m label,

     + intermediate nodes/routers are the LSRs making their forwarding
       decision on a level m label




3. Introduction

   MPLS architecture [RFC3031] defines the concept of explicitly routed
   tunnel as follows:

      If a Tunneled Packet travels from Ru to Rd over a path other
      than the Hop-by-hop path, we say that it is in an "Explicitly
      Routed Tunnel"

   where Ru and Rd are Label Switch Routers (LSRs).

   To realize explicitly routed tunnels [RFC3031] proposes to use
   explicitly routed Label Switched Paths (LSPs):

      An "Explicitly Routed LSP Tunnel" is a LSP Tunnel that is also an
      Explicitly Routed LSP

   Up until now there have been two possible protocols to instantiate/signal
   such explicitly routed LSPs - RSVP-TE ([RFC3209]) and CR-LDP
   ([RFC3212]).

   MPLS architecture ([RFC3031]) defines the notion of LSP hierarchy,
   as LSP tunnels within LSPs. Use of MPLS label stack mechanism allows
   LSP hierarchy to nest to any depth.

   In this document we specify the procedures to realize explicitly
   routed point-to-point tunnels by using LSP hierarchy, thus defining
   yet another possible mechanism to realize such tunnels. (Note though
   that the idea of using LSP hierarchy to realize explicitly routed
   tunnels is not new - e.g., Remote LFA [R-LFA] uses explicitly routed
   tunnels constructed by LSP hierarchy.)




Gredler                                                         [Page 4]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   An essential part of MPLS is the notion of label distribution
   protocol. On the subject of whether it should be one, or more then
   one label distribution protocol, MPLS architecture ([RFC3031]) said
   the following:

      THE ARCHITECTURE DOES NOT ASSUME THAT THERE IS ONLY A SINGLE
      LABEL DISTRIBUTION PROTOCOL. In fact, a number of different
      label distribution protocols are being standardized.

   Up until now IETF standardized the following label distribution
   protocols for unicast: LDP ([RFC5036]), CR-LDP ([RFC3212]), RSVP-TE
   [RFC3209] and BGP ([RFC3107], [RFC4364], [RFC4761]).

   Recently there have been proposals ([gredler-isis], [gredler-ospf],
   [previdi-isis], [psenak-ospf]) to extend IS-IS [RFC1142] and OSPF
   [RFC1583] to make them yet another label distribution protocols.

   This document describes how use of IS-IS or OSPF as label
   distribution protocols fits into the MPLS architecture. This document
   also describes the benefits of using IS-IS/OSPF as label distribution
   protocols for the purpose of constructing explicitly routed tunnels
   with stacked LSPs.


4. Constructing Explicitly Routed Tunnels by using Stacked LSPs

   Instead of explicitly routed LSPs, one can use LSP hierarchy (stack
   of LSPs) to construct explicitly routed point-to-point tunnels as
   follows.

   Consider an explicitly routed point-to-point tunnel with an explicit
   route <R(0), R(1), R(2), ... R(n)>, where R(0) is the ingress of the
   tunnel and R(n) is the egress of the tunnel. Denote the LSPs needed
   to realize such a tunnel via an LSP stack as <LSP(1), LSP(2), ...
   LSP(n)>, where LSP(1) is the topmost and LSP(n) is the bottommost LSP
   in the stack. These LSPs are constructed as follows:

     + All the LSPs in the stack are constructed with the same ingress -
       R(0).  (See further down on why this is needed.)

     + LSP(i) is constructed with R(i) as its egress (e.g., LSP(1) is
       constructed with R(1) as its egress, LSP(2) with R(2) as its
       egress, etc...  LSP(n) with R(n) as its egress).

     + For every 0 < i < n, the first intermediate router of LSP(i+1).
       is constructed to be the same as the egress router of LSP(i).





Gredler                                                         [Page 5]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


     + The first intermediate router of LSP(1) is constructed to be one
       hop away from R(0). If R(1) is one hop away from R(0), then this
       intermediate router is also the egress of LSP(1) (in which case
       LSP(1) is a one-hop LSP). If R(1) is more than one hop away from
       R(0), then this intermediate router is some router other than
       R(1), and R(1) is still the egress of that LSP.

     + The first intermediate router of any LSP in the stack, could be
       either single or multi-hop away from the egress of that LSP.

     + All the LSPs in the stack are constructed with the penultimate
       hop popping. That is, for each LSP in the stack the penultimate
       router of that LSP pops the label corresponding to that LSP off
       the label stack before sending data to the egress router of that
       LSP.


   When R(i) and R(i+1) are single hop away from each other, the first
   intermediate router of LSP(i+1) is one hop away from the egress of
   that LSP. When R(i) and R(i+1) are multi-hop away from each other,
   the first intermediate router of LSP(i+1) is multi-hop away from the
   egress of that LSP.

   Following the above procedures, the LSPs in the stack satisfy the
   following properties:

     + All LSPs in the stack have the same ingress.

     + The egress of a given LSP in the stack is the first intermediate
       router of the next LSP in the stack.

     + The first intermediate router of the LSP at the top of the stack
       is one hop away from the ingress.

     + The first intermediate router of any LSP in the stack could be
       either single or multi-hop away from the egress of that LSP.
       (Thus the egress of a given LSP in the stack could be either
       single or multi-hop away from the egress of the next LSP in the
       stack.)


   Such stack of LSPs provides the functionality to forward a packet
   through a sequence of egresses of the LSPs on the stack - the
   sequence of these egresses represents the explicit route of the
   explicitly routed point-to-point tunnel constructed by using these
   stacked LSPs. The ingress of all these LSPs is the ingress of the
   tunnel.




Gredler                                                         [Page 6]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   When the first intermediate router of a given LSP in the stack is
   multi-hop away from the egress of that LSP, the existing label
   distribution protocols (LDP, RSVP-TE, etc. ) can be used to establish
   a multi-hop LSP fragment for this LSP. When IS-IS or OSPF, in
   addition to being a routing protocol, is also used as a label
   distribution protocol (see section "IS-IS or OSPF as Label
   Distribution Protocol"), it can also be used to establish such multi-
   hop LSP fragment.

   To construct the label stack associated with the stack of LSPs the
   ingress of all these LSPs, R0, uses the following procedures:

     + For (n > i > 0) R(0) obtains from R(i) label binding for LSP(i+1)
       and places the label onto the stack, starting from the bottommost
       label (the label that corresponds to LSP(n).

       In section "IS-IS or OSPF as Label Distribution Protocol" we
       describe how IS-IS or OSPF with appropriate extensions could be
       used as a label distribution protocol to obtain such label
       bindings.

       If the first intermediate router of LSP(i+1) is either (a) a
       single hop away from the egress of that LSP, or (b) multi-hop
       away, and LDP is used as a label distribution protocol to
       establish a multi-hop LSP fragment between the first intermediate
       router and the egress of that LSP, then R(0) can use targeted LDP
       session with R(i) to obtain such label bindings.

     + For LSP(1) if R(1) is one hop away from R(0), then no label is
       needed (as LSP(1), just like all other LSPs in the stack, is
       constructed with the penultimate hop popping), and the label
       stack construction terminates with the topmost label that R(0)
       obtains from R(1) for LSP(2).

     + Otherwise, if R(1) is more than one hop away from R(0), then R(0)
       obtains label binding for LSP(1) from the first intermediate
       router of LSP(1), and places this label at the top of the stack.


   Note that the above procedures require all the LSPs in the stack to
   have the same ingress - R(0). This requirement comes from the
   observation that (a) R(0) is the router that constructs the whole
   label stack needed to realize the explicitly routed tunnel, and (b)
   according to MPLS architecture [RFC3031] when a router wants to
   create a label stack, the router has to be the head-end of all the
   LSPs corresponding to the labels in the stack.





Gredler                                                         [Page 7]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   Since the MPLS label stack mechanism allows stack of LSPs to nest to
   any depth, use of LSP hierarchy for explicitly routed tunnels does
   not place any protocol restrictions on the number of entries in the
   explicit route of an explicitly routed tunnel. Note though that there
   may be some other restrictions (e.g., due to MTU, or hardware) that
   would place an upper bound on the depth of the label stack, and thus
   on the number of entries in the explicit route.  Also, the depth of
   the label stack may have implications on ECMP, and specifically on
   the use of the Entropy label (see [kini] for more).



4.1. Examples of Constructing Explicitly Routed Tunnels by Stacked LSPs

   In this section we illustrate how to construct an explicitly routed
   tunnel by using stacked LSPs. The first example illustrates this for
   an explicitly routed tunnel where consecutive hops that define the
   tunnel are one hop away from each other. The second example
   illustrates this when these hops are more than one hop away from each
   other.


4.1.1. Explicitly Routed Tunnel with Single Hops

   Consider a network topology shown below:

      R0-----R4
      |     /|
      |    / |
      |   /  |
      |  /   |
      | /    |
      R1    R3
       \    /
        \  /
         R2


   Assume that R0 wants to construct an explicitly routed tunnel with
   (R0, R1, R2, R3, R4). The consecutive hops that define the tunnel are
   one hop away from each other. That is, R0 is one hop away from R1, R2
   is one hop away from R1, R3 is one hop away from R2, and R4 is one
   hop away from R3.

   R0 constructs this tunnel using the following stack of LSPs:

      LSP1: (R0, R1) - top of the stack
      LSP2: (R0, R1, R2)



Gredler                                                         [Page 8]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


      LSP3: (R0, R2, R3)
      LSP4: (R0, R3, R4) - bottom of the stack

   Note that this stack of LSPs meets the requirements specified in
   section "Constructing Explicitly Routed Tunnels by using Stacked
   LSPs". Specifically,

     + All four LSPs in the stack have the same ingress - R0, which is
       also the ingress of the explicitly routed tunnel.

     + The egress of LSP1, R1, is the first intermediate router of the
       next LSP in the stack, LSP2. The egress of LSP2, R2, is the first
       intermediate router of the next LSP in the stack, LSP3. Likewise,
       the egress of LSP3, R3, is the first intermediate router of the
       next (and the last) LSP in the stack, LSP4.

     + The LSP at the top of the stack, LSP1, has its first intermediate
       router, R1, one hop away from its ingress, R0. Because of that,
       this intermediate router is also the egress of that LSP, and that
       LSP is a one-hop LSP.

     + In that particular example the first intermediate router of every
       LSP in the stack is one hop away from the egress of that LSP.
       That is, the first intermediate router of LSP2, R1, is one hop
       away from the egress of that LSP, R2; the first intermediate
       router of LSP3, R2, is one hop away from the egress of that LSP,
       R3; and the first intermediate router of LSP4, R3, is one hop
       away from the egress of that LSP, R4.  As a result, the egress of
       a given LSP in the stack is one hop away from the egress of the
       next LSP in the stack.


   The first intermediate router of each of these LSPs creates label
   bindings for these LSPs as follows. R3 creates label binding for LSP4
   by binding a particular label, L1, to the address of R4, creating a
   Next Hop Label Forwarding Entry (NHLFE) whose next hop is the link
   from R3 to R4, and setting the Incoming Label Map (ILM) so that L1
   maps to that NHLFE. Likewise, R2 creates label binding for LSP3 by
   binding a particular label, L2, to the address of R3, creating an
   NHLFE whose next hop is the link from R2 to R3, and setting the ILM
   so that L2 maps to that NHLFE.  Finally, R1 creates label binding for
   LSP2 by binding a particular label, L3, to the address of R2,
   creating an NHLFE whose next hop is the link from R1 to R2, and
   setting the ILM so that L3 maps to that NHLFE.

   To get from the first hop of LSP4, R0, to the second hop of LSP4, R3,
   the packet has to go through the LSP tunnel provided by LSP3.  To get
   from the first hop of LSP3, R0, to the second hop of LSP3, R2, a



Gredler                                                         [Page 9]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   packet has to go through the LSP tunnel provided by LSP2. To get from
   the first hop of LSP2, R0, to the second hop of LSP2, R1, a packet
   has to go through the LSP tunnel provided by LSP1.

   In order to accomplish this R0 constructs the label stack for the
   explicitly routed tunnel as follows:

     + Step 1: R0 obtains label binding L1 created by R3 for LSP4 (R0,
       R3, R4), and starts building the label stack by pushing L1 onto
       the label stack.

     + Step 2: R0 obtains label binding L2 created by R2 for LSP3 (R0,
       R2, R3), and  pushes L2 into the stack. At this point the stack
       contains (L2, L1).

     + Step 3: R0 obtains label binding L3 created by R1 for LSP2 (R0,
       R1, R2), and pushes L3 into the stack. At this point the stack
       contains (L3, L2, L1).

     + Step 4: Since R0 and R1 are one hop away from each other the
       label stack construction is completed (R0 does not need a label
       for one-hop LSP1, as all the LSPs use penultimate hop popping).


   So far we did not say anything about how R0 obtains from R3 label
   binding for LSP4, from R2 label binding for LSP3, and from R1 label
   binding for LSP2.

   At least in principle, these label bindings could be obtain by such
   already defined label distribution protocols as LDP (to be more
   precise, targeted LDP if the two routers are more than one hop away
   from each other). E.g., if one uses targeted LDP, then R0 would need
   to dynamically establish and maintain a targeted LDP session with R3
   and another targeted LDP session with R2 (R0 would maintain a
   "vanilla" LDP session with R1). Using these LDP sessions R0 would
   obtain from R3 label binding for LSP4, from R2 label binding for
   LSP3, and from R1 label binding for LSP2. Note that obtaining such
   labels bindings with targeted LDP may also require defining a new FEC
   to be used by targeted LDP.

   In section "IS-IS or OSPF as Label Distribution Protocol" we describe
   how IS-IS or OSPF with appropriate extensions could be used as a
   label distribution protocol to obtain such label bindings.

   When R0 wants to forward a packet along the explicitly constructed
   tunnel (R0, R1, R2, R3, R4), R0 pushes (L3, L2, L1) onto the label
   stack of the packet, and forwards the packet to R1. R1 performs the
   lookup on the topmost label, L3, and based on this lookup forwards



Gredler                                                        [Page 10]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   the packet to R2. Prior to forwarding the packet to R2, R1 (acting as
   a penultimate hop for LSP2) pops the topmost label, L3.  When R2
   receives the packet, R2 performs the lookup on the topmost label, L2,
   and based on this lookup forwards the packet to R3. Prior to
   forwarding the packet to R3, R2 (acting as a penultimate hop for
   LSP3) pops the topmost label, L2. When R3 receives the packet, R3
   performs the lookup on the topmost label, L1, and based on this
   lookup forwards the packet to R4. Prior to forwarding the packet to
   R4, R3 (acting as a penultimate hop for LSP4) pops the topmost label,
   L1.



4.1.2. Explicitly Routed Tunnel with Multi-Hops

   Consider a network topology shown below:

         R0---R1
         /     \
        R2      R5---R6
         \     /
         R3---R4

   Assume that R0 wants to construct an explicitly routed tunnel with
   (R0, R4, R6) as hops. Note that R4 is multi-hop away from R0, and R6
   is multi-hop away from R4.

   R0 constructs this tunnel using the following stack of LSPs:

      LSP1: (R0, R2, R3, R4) - top of the stack
      LSP2: (R0, R4, R5, R6) - bottom of the stack

   Note that this stack of LSPs meets the requirements specified in
   section "Constructing Explicitly Routed Tunnels by using Stacked
   LSPs". Specifically,

     + Both LSPs in the stack have the same ingress - R0, which is also
       the ingress of the explicitly routed tunnel.

     + The egress of LSP1, R4, is the first intermediate router of the
       next LSP in the stack, LSP2.

     + The LSP at the top of the stack, LSP1, has its first intermediate
       router, R2, one hop away from its ingress, R0. However, this
       intermediate router is not the egress of that LSP, and therefore
       this LSP is a multi-hop LSP.





Gredler                                                        [Page 11]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


     + In that particular example the first intermediate router of every
       LSP in the stack is multi-hop away from the egress of that LSP.
       That is, the first intermediate router of LSP1, R2, is multi-hop
       away from the egress of that LSP, R4; the first intermediate
       router of LSP2, R4, is multi-hop away from the egress of that
       LSP, R6.  As a result, the egress of a given LSP in the stack is
       multi-hop away from the egress of the next LSP in the stack.


   In this example we assume that LDP is used as a label distribution
   protocol for both LSP1 and LSP2. Since R0 and R4 are not IGP
   neighbors, they are remote label distribution peers. Thus R0 and R4
   use targeted LDP for label distribution. All other routers use
   "vanilla" LDP procedures.

   To get from the first hop of LSP2, R0, to its second hop, R4, the
   packet has to go through the LSP tunnel provided by LSP1.

   In order to accomplish this R0 constructs the label stack for the
   explicitly routed tunnel as follows:

     + Step 1: R0 (using targeted LDP) obtains label binding L1 created
       by R4 for LSP2 (R0, R4, R5, R6), and starts building the label
       stack by pushing L1 onto the label stack.

     + Step 2: R0 (using "vanilla" LDP procedures) obtains label binding
       L2 created by R2 for LSP1 (R0, R2, R3, R4), and  pushes L2 into
       the stack. At this point the stack contains (L2, L1).

     + Step 3: Since R0 and R1 are one hop away from each other the
       label stack construction is completed (R0 does not need a label
       for one-hop LSP1).


   A reader familiar with Remote LFA FRR [R-LFA] should be able to
   notice that the example described in this section is nothing more
   than an instance of Remote LFA FRR, where Remote LFA FRR provides
   fast reroute to the traffic going from R0 to R6 in the presence of
   the (R0, R2) link failure, with R0 being the Point of Local Repair
   (PLR), R4 being the PQ-node, and R6 being the ultimate destination.
   The explicitly routed tunnel (R0, R4, R6) consists of the PLR as the
   head-node, the PQ-node as the next hop, and the ultimate destination
   as yet another hop.








Gredler                                                        [Page 12]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


5. IS-IS or OSPF as Label Distribution Protocol

   When OSPF or IS-IS, in addition to being a routing protocol, is also
   used as a label distribution protocol (as proposed in [gredler-isis],
   [gredler-ospf], [previdi-isis], [psenak-ospf), the OSPF/IS-IS Link
   State Advertisements originated by a router carry label bindings for
   LSPs that either transit or originated by the router. Doing this
   allows to extend such LSPs. The criteria for selecting among all
   these LSPs a subset for which the router would originate label
   binding advertisements in IS-IS/OSPF are purely local to the router.
   The router could be either single or multi-hop away from the egresses
   of the LSPs in the subset. Existing label distribution protocols
   (LDP, RSVP-TE, etc.) can be used to establish multi-hop LSP fragments
   if the router is multi-hop away from the egress of a particular LSP
   in the subset. When IS-IS or OSPF, in addition to being a routing
   protocol, is also used as a label distribution protocol, it can also
   be used to establish such multi-hop LSP fragments.

   Use of IS-IS or OSPF as a label distribution protocol supports
   advertisements of label mappings for such FECs as:

     + IPv4/IPv6 prefix FEC via a hop-by-hop LSP established using IS-
       IS/OSPF as a label distribution protocol.

     + IPv4/IPv6 prefix FEC via a hop-by-hop LSP established using LDP
       as a label distribution protocol.

     + IPv4/IPv6 address FEC where the address identifies the remote end
       of one of the advertising router's links via an LSP that
       traverses the link and terminates on the remote end of the link.

     + IPv4/IPv6 address FEC where the address identifies the remote end
       of one of the advertising router's point-to-point links via an
       LSP that traverses the link and terminates on the remote end of
       the link.

     + IPv4/IPv6 address FEC where the address identifies a (remote)
       router connected to the advertising router by a broadcast link
       via an LSP that traverses the link and terminates on the remote
       router identified by its node-id.

     + IPv4/IPv6 prefix FEC via an explicitly routed LSP established
       using RSVP-TE, where path computation for such LSP is done by
       either distributed CSPF, or by PCE.


   When a router obtains label binding for a given FEC from more than
   one label distribution protocol (e.g., one binding from targeted LDP



Gredler                                                        [Page 13]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   and another from IS-IS/OSPF), deciding which label binding to use is
   a matter of policy local to the router. In the scenario where a
   router obtains label binding for a given FEC from both (targeted) LDP
   and IS-IS/OSPF, the default behavior RECOMMENDED in this document is
   to prefer the one obtained from (targeted) LDP.  An implementation
   MUST support the ability to override the default behavior via
   configuration.

   MPLS architecture [RFC3031] defines the notion of local/remote label
   distribution peers as follows:

      When two LSRs are IGP neighbors, we will refer to them as "local
      label distribution peers".  When two LSRs may be label distribution
      peers, but are not IGP neighbors, we will refer to them as "remote
      label distribution peers."

   Following OSPF/IS-IS procedures each router passes Link State
   Advertisements originated by other routers unmodified. When these
   advertisements carry label binding information, this information is
   also passed unmodified. Therefore, the router that originates label
   bindings advertisements in IS-IS/OSPF can be either single or multi-
   hop away from the routers that receive and use these bindings.  In
   the former case the IGP neighbors of the router that originates the
   advertisements will be the local label distribution peers of the
   router. In the latter case other routers in the same IGP domain will
   be the remote label distribution peers of the router.

   Use of OSPF or IS-IS as a label distribution protocol provides
   scalable support for remote label distribution peering in terms of
   the number of label distribution peers a given router has to
   maintain.  This is because label distribution protocol messages (Link
   State Advertisements) are exchanged only between IGP neighbors,
   without requiring control plane peering between a router that
   originates Link State Advertisements and each of its remote label
   distribution peers.

   It is important to note that the existing MPLS control plane already
   has mechanisms/protocols to support remote label distribution peering
   (using BGP or targeted LDP [RFC5036]). Thus the practical relevance
   of the ability to provide scalable support for remote label
   distribution peering with IS-IS or OSPF as a label distribution
   protocol depends on a particular use case.

   If for a given subset of routers within an MPLS network each router
   within the subset is assigned a distinct index, then one could
   compress announcements of labels bound to the LSPs whose FECs are the
   IP addresses of these routers by (a) advertising these indices in IS-
   IS/OSPF, and (b) making each router advertise a label block in IS-



Gredler                                                        [Page 14]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   IS/OSPF as well. A router R1 that advertises a given label block
   algorithmically binds a FEC associated with an IP address of some
   other router R2 to the label from that block that is identified by
   the index that R2 advertises in IGP. A router R1 that receives label
   block originated by some other router R2 can determine the label
   bound to a FEC associated with an IP address of some other router R3
   by using the index advertised by R3 as an offset into the label block
   advertised by R2. Note that to avoid wasting labels this scheme
   requires a fairly dense assignment of indices. Also note that to
   expand the number of labels that a router advertises using label
   blocks, the router may advertise more than one label block.

   Note though, that the benefits of scaling improvements in terms of
   label distribution peering come at a cost, as every router in the
   domain ends up keeping all the labels assigned/bounded by every other
   router in the domain, whether it really needs to know them or not.
   Whether this cost is of practical significance depends on (a) the
   number of label bindings being advertised, and (b) the encoding of
   label bindings (e.g., use of label blocks vs enumerating each label
   binding).



5.1. Example of IS-IS/OSPF as Label Distribution Protocols

   In this section we illustrate how IS-IS/OSPF with extensions, as
   defined in [gredler-isis], [gredler-ospf], [previdi-isis], [psenak-
   ospf] could be used as a label distribution protocol to support
   explicitly routed tunnels realized by stacked LSPs. For the purpose
   of this illustration we assume the scenario described in section
   "Example of Constructing Explicitly Routed Tunnels by Stacked LSPs".
   In that example one of the key issues is the ability of R0 to obtain
   from R3 label binding for LSP4, from R2 label binding for LSP3, and
   from R1 label binding for LSP2.

   To obtains such label bindings, the Link State Advertisement
   originated by R3 carries label L1 (this is the label that R3 binds to
   LSP4). Using IS-IS/OSPF procedures this Link State Advertisement is
   propagated by R2 and R1 (as well as by R4) to R0. This is how R0
   obtains from R3 label binding for LSP4. In a similar fashion, the
   Link State Advertisement originated by R2 carries label L2 (which is
   the label that R2 binds to LSP3). Using IS-IS/OSPF procedures this
   Link State Advertisement is propagated by R1 (as well as by R3 and
   R4) to R0. This is how R0 obtains from R2 label binding for LSP3.
   Likewise, the Link State Advertisement originated by R1 carries label
   L3 (which is the label that R1 binds to LSP2).  Using IS-IS/OSPF
   procedures this Link State Advertisement is delivered to R0. This is
   how R0 obtains from R1 label binding for LSP2.



Gredler                                                        [Page 15]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   Note that while from R0's perspective both R2 and R3 are remote label
   distribution peers, R0 does not maintain any control plane peering
   (e.g., targeted LDP) with either R2 or R3.


6. IANA Considerations

   This document introduces no new IANA Considerations.


7. Security Considerations

   TBD


8. Acknowledgements

   We would like to thank John Drake (Juniper Networks) and John Scudder
   (Juniper Networks) for their review and comments.

   We would also like to thank Bruno Decraene (Orange) and Robert Raszuk
   for their review and comments.


9. Normative References

   [RFC3031] Rosen, E., et. al., "Multiprotocol Label Switching
   Architecture", RFC3031, January 2001


10. Informative References

   [kini] Kini, S., et. al., "Entropy labels for source routed stacked
   tunnels", draft-kini-mpls-entropy-label-src-stacked-tunnels, work in
   progress

   [gredler] Gredler, H., et. al., "Advertising MPLS labels in IGPs"
   draft-gredler-rtgwg-igp-label-advertisement, work in progress

   [gredler-isis] Gredler, H., et. al., "Advertising MPLS labels in IS-
   IS" draft-gredler-isis-label-advertisement, work in progress

   [gredler-ospf] Gredler, H., et. al., "Advertising MPLS labels in
   OSPF" draft-gredler-ospf-label-advertisement, work in progress

   [previdi-isis] Previdi, S., et. al., "IS-IS Extensions for Segment
   Routing" draft-previdi-isis-segment-routing-extensions, work in
   progress



Gredler                                                        [Page 16]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   [psenak-ospf] Psenak, P., et. al., "OSPF Extensions for Segment
   Routing" draft-psenak-ospf-segment-routing-extensions, work in
   progress

   [R-LFA] Bryant, S., et. al., "Remote LFA FRR", draft-ietf-rtgwg-
   remote-lfa, work in progress

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

   [RFC3107] Rekhter, Y., et. al., "Carrying Label Information in
   BGP-4", RFC3107, May 2001

   [RFC3209] Awduche, D., et. al., "RSVP-TE: Extensions to RSVP for LSP
   Tunnels", RFC3209, December 2001

   [RFC3212] Jamoussi, B., et. al., "Constraint-Based LSP Setup using
   LDP", RFC3212, January 2002

   [RFC4364] Rosen E., et. al., "BGP/MPLS IP Virtual Private Networks
   (VPNs)", RFC4364, February 2006

   [RFC4761] Kompella, K., et. al., "Virtual Private LAN Service (VPLS)
   Using BGP for Auto-Discovery and Signaling", RFC4761, January 2007

   [RFC5036] L. Andersson, et. al., "LDP Specification", RFC5036,
   October 2007


11. Authors' Addresses

   Hannes Gredler
   Juniper Networks
   e-mail: hannes@juniper.net

   Yakov Rekhter
   Juniper Networks
   e-mail: yakov@juniper.net

   Luay Jalil
   Verizon
   e-mail: luay.jalil@verizon.com

   Sriganesh Kini
   Ericsson
   Email: sriganesh.kini@ericsson.com

   Xiaohu Xu



Gredler                                                        [Page 17]


Internet Draft      draft-gredler-spring-mpls-06.txt            May 2014


   Huawei Technologies,
   Email: xuxiaohu@huawei.com

















































Gredler                                                        [Page 18]


Html markup produced by rfcmarkup 1.129c, available from https://tools.ietf.org/tools/rfcmarkup/