path: root/doc/rfc/rfc3906.txt

Network Working Group                                            N. Shen
Request for Comments: 3906                              Redback Networks
Category: Informational                                          H. Smit
                                                            October 2004


          Calculating Interior Gateway Protocol (IGP) Routes
                    Over Traffic Engineering Tunnels

Status of this Memo

   This memo provides information for the Internet community.  It does
   not specify an Internet standard of any kind.  Distribution of this
   memo is unlimited.

Copyright Notice

   Copyright (C) The Internet Society (2004).

Abstract

   This document describes how conventional hop-by-hop link-state
   routing protocols interact with new Traffic Engineering capabilities
   to create Interior Gateway Protocol (IGP) shortcuts.  In particular,
   this document describes how Dijkstra's Shortest Path First (SPF)
   algorithm can be adapted so that link-state IGPs will calculate IP
   routes to forward traffic over tunnels that are set up by Traffic
   Engineering.

1.  Introduction

   Link-state protocols like integrated Intermediate System to
   Intermediate System (IS-IS) [1] and OSPF [2] use Dijkstra's SPF
   algorithm to compute a shortest path tree to all nodes in the
   network.  Routing tables are derived from this shortest path tree.
   The routing tables contain tuples of destination and first-hop
   information.  If a router does normal hop-by-hop routing, the first-
   hop will be a physical interface attached to the router.  New traffic
   engineering algorithms calculate explicit routes to one or more nodes
   in the network.  At the router that originates explicit routes, such
   routes can be viewed as logical interfaces which supply Label
   Switched Paths through the network.  In the context of this document,
   we refer to these Label Switched Paths as Traffic Engineering tunnels
   (TE-tunnels).  Such capabilities are specified in [3] and [4].

   The existence of TE-tunnels in the network and how the traffic in the
   network is switched over those tunnels are orthogonal issues.  A node
   may define static routes pointing to the TE-tunnels, it may match the



Shen & Smit                  Informational                      [Page 1]
^L
RFC 3906              IGP ShortCut Over MPLS LSPs           October 2004


   recursive route next-hop with the TE-tunnel end-point address, or it
   may define local policy such as affinity based tunnel selection for
   switching certain traffic.  This document describes a mechanism
   utilizing link-state IGPs to dynamically install IGP routes over
   those TE-tunnels.

   The tunnels under consideration are tunnels created explicitly by the
   node performing the calculation, and with an end-point address known
   to this node.  For use in algorithms such as the one described in
   this document, it does not matter whether the tunnel itself is
   strictly or loosely routed.  A simple constraint can ensure that the
   mechanism be loop free.  When a router chooses to inject a packet
   addressed to a destination D, the router may inject the packet into a
   tunnel where the end-point is closer (according to link-state IGP
   topology) to the destination D than is the injecting router.  In
   other words, the tail-end of the tunnel has to be a downstream IGP
   node for the destination D.  The algorithms that follow are one way
   that a router may obey this rule and dynamically make intelligent
   choices about when to use TE-tunnels for traffic.  This algorithm may
   be used in conjunction with other mechanisms such as statically
   defined routes over TE-tunnels or traffic flow and QoS based TE-
   tunnel selection.

   This IGP shortcut mechanism assumes the TE-tunnels have already been
   setup.  The TE-tunnels in the network may be used for QoS, bandwidth,
   redundancy, or fastreroute reasons.  When an IGP shortcut mechanism
   is applied on those tunnels, or other mechanisms are used in
   conjunction with an IGP shortcut, the physical traffic switching
   through those tunnels may not match the initial traffic engineering
   setup goal.  Also the traffic pattern in the network may change with
   time.  Some forwarding plane measurement and feedback into the
   adjustment of TE-tunnel attributes need to be there to ensure that
   the network is being traffic engineered efficiently [6].

2.  Enhancement to the Shortest Path First Computation

   During each step of the SPF computation, a router discovers the path
   to one node in the network.  If that node is directly connected to
   the calculating router, the first-hop information is derived from the
   adjacency database.  If a node is not directly connected to the
   calculating router, it inherits the first-hop information from the
   parent(s) of that node.  Each node has one or more parents.  Each
   node is the parent of zero or more down-stream nodes.

   For traffic engineering purposes, each router maintains a list of all
   TE-tunnels that originate at this router.  For each of those TE-
   tunnels, the router at the tail-end is known.




Shen & Smit                  Informational                      [Page 2]
^L
RFC 3906              IGP ShortCut Over MPLS LSPs           October 2004


   During SPF, when a router finds the path to a new node (in other
   words, this new node is moved from the TENTative list to the PATHS
   list), the router must determine the first-hop information.  There
   are three possible ways to do this:

      -  Examine the list of tail-end routers directly reachable via a
         TE-tunnel.  If there is a TE-tunnel to this node, we use the
         TE-tunnel as the first-hop.

      -  If there is no TE-tunnel, and the node is directly connected,
         we use the first-hop information from the adjacency database.

      -  If the node is not directly connected, and is not directly
         reachable via a TE-tunnel, we copy the first-hop information
         from the parent node(s) to the new node.

   The result of this algorithm is that traffic to nodes that are the
   tail-end of TE-tunnels, will flow over those TE-tunnels.  Traffic to
   nodes that are downstream of the tail-end nodes will also flow over
   those TE-tunnels.  If there are multiple TE-tunnels to different
   intermediate nodes on the path to destination node X, traffic will
   flow over the TE-tunnel whose tail-end node is closest to node X.  In
   certain applications, there is a need to carry both the native
   adjacency and the TE-tunnel next-hop information for the TE-tunnel
   tail-end and its downstream nodes.  The head-end node may
   conditionally switch the data traffic onto TE-tunnels based on user
   defined criteria or events; the head-end node may also split flow of
   traffic towards either types of the next-hops; the head-end node may
   install the routes with two different types of next-hops into two
   separate RIBs.  Multicast protocols running over physical links may
   have to perform RPF checks using the native adjacency next-hops
   rather than the TE-tunnel next-hops.

3.  Special Cases and Exceptions

   The Shortest Path First algorithm will find equal-cost parallel paths
   to destinations.  The enhancement described in this document does not
   change this.  Traffic can be forwarded over one or more native IP
   paths, over one or more TE-tunnels, or over a combination of native
   IP paths and TE-tunnels.

   A special situation occurs in the following topology:

      rtrA -- rtrB -- rtrC
               |       |
              rtrD -- rtrE





Shen & Smit                  Informational                      [Page 3]
^L
RFC 3906              IGP ShortCut Over MPLS LSPs           October 2004


   Assume all links have the same cost.  Assume a TE-tunnel is set up
   from rtrA to rtrD.  When the SPF calculation puts rtrC on the
   TENTative list, it will realize that rtrC is not directly connected,
   and thus it will use the first-hop information from the parent, which
   is rtrB.  When the SPF calculation on rtrA moves rtrD from the
   TENTative list to the PATHS list, it realizes that rtrD is the tail-
   end of a TE-tunnel.  Thus rtrA will install a route to rtrD via the
   TE-tunnel, and not via rtrB.

   When rtrA puts rtrE on the TENTative list, it realizes that rtrE is
   not directly connected, and that rtrE is not the tail-end of a TE-
   tunnel.  Therefore, rtrA will copy the first-hop information from the
   parents (rtrC and rtrD) to the first-hop information of rtrE.
   Traffic to rtrE will now load-balance over the native IP path via
   rtrA->rtrB->rtrC, and the TE-tunnel rtrA->rtrD.

   In the case where both parallel native IP paths and paths over TE-
   tunnels are available, implementations can allow the network
   administrator to force traffic to flow over only TE-tunnels (or only
   over native IP paths) or both to be used for load sharing.

4.  Metric Adjustment of IP Routes over TE-tunnels

   When an IGP route is installed in the routing table with a TE-tunnel
   as the next hop, an interesting question is what should be the cost
   or metric of this route?  The most obvious answer is to assign a
   metric that is the same as the IGP metric of the native IP path as if
   the TE-tunnels did not exist.  For example, rtrA can reach rtrC over
   a path with a cost of 20.  X is an IP prefix advertised by rtrC.  We
   install the route to X in rtrA's routing table with a cost of 20.
   When a TE-tunnel from rtrA to rtrC comes up, by default the route is
   still installed with metric of 20, only the next-hop information for
   X is changed.

   While this scheme works well, in some networks it might be useful to
   change the cost of the path over a TE-tunnel, to make the route over
   the TE-tunnel less or more preferred than other routes.

   For instance, when equal cost paths exist over a TE-tunnel and over a
   native IP path, by adjusting the cost of the path over the TE-tunnel,
   we can force traffic to prefer the path via the TE-tunnel, to prefer
   the native IP path, or to load-balance among them.  Another example
   is when multiple TE-tunnels go to the same or different destinations.
   Adjusting TE-tunnel metrics can force the traffic to prefer some TE-
   tunnels over others regardless of underlining IGP cost to those
   destinations.





Shen & Smit                  Informational                      [Page 4]
^L
RFC 3906              IGP ShortCut Over MPLS LSPs           October 2004


   Setting a manual metric on a TE-tunnel does not impact the SPF
   algorithm itself.  It only affects the comparison of the new route
   with existing routes in the routing table.  Existing routes can be
   either IP routes to another router that advertises the same IP
   prefix, or it can be a path to the same router, but via a different
   outgoing interface or different TE-tunnel.  All routes to IP prefixes
   advertised by the tail-end router will be affected by the TE-tunnel
   metric.  Also, the metrics of paths to routers that are downstream of
   the tail-end router will be influenced by the manual TE-tunnel
   metric.

   This mechanism is loop free since the TE-tunnels are source-routed
   and the tunnel egress is a downstream node to reach the computed
   destinations.  The end result of TE-tunnel metric adjustment is more
   control over traffic loadsharing.  If there is only one way to reach
   a particular IP prefix through a single TE-tunnel, then no matter
   what metric is assigned, the traffic has only one path to go.

   The routing table described in this section can be viewed as the
   private RIB for the IGP.  The metric is an important attribute to the
   routes in the routing table.  A path or paths with lower metric will
   be selected over other paths for the same route in the routing table.

4.1.  Absolute and Relative Metrics

   It is possible to represent the TE-tunnel metric in two different
   ways: an absolute (or fixed) metric or a relative metric, which is
   merely an adjustment of the dynamic IGP metric as calculated by the
   SPF computation.  When using an absolute metric on a TE-tunnel, the
   cost of the IP routes in the routing table does not depend on the
   topology of the network.  Note that this fixed metric is not only
   used to compute the cost of IP routes advertised by the router that
   is the tail-end of the TE-tunnel, but also for all the routes that
   are downstream of this tail-end router.  For example, if we have TE-
   tunnels to two core routers in a remote POP, and one of them is
   assigned with an absolute metric of 1, then all the traffic going to
   that POP will traverse this low-metric TE-tunnel.

   By setting a relative metric, the cost of IP routes in the routing
   table is based on the IGP metric as calculated by the SPF
   computation.  This relative metric can be a positive or a negative
   number.  Not configuring a metric on a TE-tunnel is a special case of
   the relative metric scheme.  No metric is the same as a relative
   metric of 0.  The relative metric is bounded by minimum and maximum
   allowed metric values while the positive metric disables the TE-
   tunnel in the SPF calculation.





Shen & Smit                  Informational                      [Page 5]
^L
RFC 3906              IGP ShortCut Over MPLS LSPs           October 2004


4.2.  Examples of Metric Adjustment

   Assume the following topology.  X, Y, and Z are IP prefixes
   advertised by rtrC, rtrD, and rtrE respectively.  T1 is a TE-tunnel
   from rtrA to rtrC.  Each link in the network has an IGP metric of 10.

        ===== T1 =====>
      rtrA -- rtrB -- rtrC -- rtrD -- rtrE
           10      10  |   10  |   10  |
                       X       Y       Z

   Without TE-tunnel T1, rtrA will install IP routes X, Y, and Z in the
   routing table with metrics 20, 30, and 40 respectively.  When rtrA
   has brought up TE-tunnel T1 to rtrC, and if rtrA is configured with
   the relative metric of -5 on tunnel T1, then the routes X, Y, and Z
   will be installed in the routing table with metrics 15, 25, and 35.
   If an absolute metric of 5 is configured on tunnel T1, then rtrA will
   install routes X, Y, and Z all with metrics 5, 15, and 25
   respectively.

5.  Security Considerations

   This document does not change the security aspects of IS-IS or OSPF.
   Security considerations specific to each protocol still apply.  For
   more information see [5] and [2].

6.  Acknowledgments

   The authors would like to thank Joel Halpern and Christian Hopps for
   their comments on this document.

7.  Informative References

   [1] ISO.  Information Technology - Telecommunications and Information
       Exchange between Systems - Intermediate System to Intermediate
       System Routing Exchange Protocol for Use in Conjunction with the
       Protocol for Providing the Connectionless-Mode Network Service.
       ISO, 1990.

   [2] Moy, J., "OSPF Version 2", RFC 2328, April 1998.

   [3] Awduche, D., Malcolm, J., Agogbua, J., O'Dell, M., and J.
       McManus, "Requirements for Traffic Engineering Over MPLS", RFC
       2702, September 1999.

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



Shen & Smit                  Informational                      [Page 6]
^L
RFC 3906              IGP ShortCut Over MPLS LSPs           October 2004


   [5] Li, T. and R. Atkinson, "Intermediate System to Intermediate
       System (IS-IS) Cryptographic Authentication", RFC 3567, July
       2003.

   [6] Awduche, D., Chiu, A., Elwalid, A., Widjaja, I., and X. Xiao,
       "Overview and Principles of Internet Traffic Engineering", RFC
       3272, May 2002.

8.  Authors' Addresses

   Naiming Shen
   Redback Networks, Inc.
   300 Holger Way
   San Jose, CA 95134

   EMail: naiming@redback.com


   Henk Smit

   EMail: hhwsmit@xs4all.nl






























Shen & Smit                  Informational                      [Page 7]
^L
RFC 3906              IGP ShortCut Over MPLS LSPs           October 2004


9.  Full Copyright Statement

   Copyright (C) The Internet Society (2004).

   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 IETF's procedures with respect to rights in IETF Documents can
   be found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at
   http://www.ietf.org/ipr.

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at ietf-
   ipr@ietf.org.

Acknowledgement

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







Shen & Smit                  Informational                      [Page 8]
^L