RFC5151

Network Working Group A. Farrel, Ed. Request for Comments: 5151 Old Dog Consulting Updates: 3209, 3473 A. Ayyangar Category: Standards Track Juniper Networks

                                                         JP. Vasseur
                                                 Cisco Systems, Inc.
                                                       February 2008

      Inter-Domain MPLS and GMPLS Traffic Engineering --
Resource Reservation Protocol-Traffic Engineering (RSVP-TE) Extensions

Status of This Memo

This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited.

Abstract

This document describes procedures and protocol extensions for the use of Resource Reservation Protocol-Traffic Engineering (RSVP-TE) signaling in Multiprotocol Label Switching-Traffic Engineering (MPLS-TE) packet networks and Generalized MPLS (GMPLS) packet and non-packet networks to support the establishment and maintenance of Label Switched Paths that cross domain boundaries.

For the purpose of this document, a domain is considered to be any collection of network elements within a common realm of address space or path computation responsibility. Examples of such domains include Autonomous Systems, Interior Gateway Protocol (IGP) routing areas, and GMPLS overlay networks.

Introduction

The requirements for inter-area and inter-AS (Autonomous System) Multiprotocol Label Switching (MPLS) Traffic Engineering (TE) are stated in RFC4105 and RFC4216, respectively. Many of these requirements also apply to Generalized MPLS (GMPLS) networks. The framework for inter-domain MPLS-TE is provided in RFC4726.

This document presents procedures and extensions to Resource Reservation Protocol-Traffic Engineering (RSVP-TE) signaling for the setup and maintenance of traffic engineered Label Switched Paths (TE LSPs) that span multiple domains in MPLS-TE or GMPLS networks. The signaling procedures described in this document are applicable to MPLS-TE packet LSPs established using RSVP-TE (RFC3209) and all LSPs (packet and non-packet) that use RSVP-TE GMPLS extensions as described in RFC3473.

Three different signaling methods for inter-domain RSVP-TE signaling are identified in RFC4726. Contiguous LSPs are achieved using the procedures of RFC3209 and RFC3473 to create a single end-to-end LSP that spans all domains. Nested LSPs are established using the techniques described in RFC4206 to carry the end-to-end LSP in a separate tunnel across each domain. Stitched LSPs are established using the procedures of RFC5150 to construct an end-to-end LSP from the concatenation of separate LSPs each spanning a domain.

This document defines the RSVP-TE protocol extensions necessary to control and select which of the three signaling mechanisms is used for any one end-to-end inter-domain TE LSP.

For the purpose of this document, a domain is considered to be any collection of network elements within a common realm of address space or path computation responsibility. Examples of such domains include Autonomous Systems, IGP areas, and GMPLS overlay networks (RFC4208).

Conventions Used in This Document

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 RFC2119.

Terminology

AS: Autonomous System.

ASBR: Autonomous System Border Router. A router used to connect together ASs of a different or the same Service Provider via one or more inter-AS links.

Bypass Tunnel: An LSP that is used to protect a set of LSPs passing over a common facility.

ERO: Explicit Route Object.

FA: Forwarding Adjacency.

LSR: Label Switching Router.

MP: Merge Point. The node where bypass tunnels meet the protected LSP.

NHOP bypass tunnel: Next-Hop Bypass Tunnel. A backup tunnel, which bypasses a single link of the protected LSP.

NNHOP bypass tunnel: Next-Next-Hop Bypass Tunnel. A backup tunnel, which bypasses a single node of the protected LSP.

PLR: Point of Local Repair. The ingress of a bypass tunnel.

RRO: Record Route Object.

TE link: Traffic Engineering link.

Signaling Overview

The RSVP-TE signaling of a TE LSP within a single domain is described in RFC3209 and RFC3473. Inter-domain TE LSPs can be supported by one of three options as described in RFC4726 and set out in the next section:

- contiguous LSPs - nested LSPs - stitched LSPs.

In fact, as pointed out in RFC4726, any combination of these three options may be used in the course of an end-to-end inter-domain LSP. That is, the options should be considered as per-domain transit options so that an end-to-end inter-domain LSP that starts in domain A, transits domains B, C, and D, and ends in domain E might use an

LSP that runs contiguously from the ingress in domain A, through domain B to the border with domain C. Domain C might be transited using the nested LSP option to reach the border with domain D, and domain D might be transited using the stitched LSP option to reach the border with domain E, from where a normal LSP runs to the egress.

This document describes the RSVP-TE signaling extensions required to select and control which of the three signaling mechanisms is used.

The specific protocol extensions required to signal each LSP type are described in other documents and are out of scope for this document. Similarly, the routing extensions and path computation techniques necessary for the establishment of inter-domain LSPs are out of scope. An implementation of a transit LSR is unaware of the options for inter-domain TE LSPs since it sees only TE LSPs. An implementation of a domain border LSR has to decide what mechanisms of inter-domain TE LSP support to include, but must in any case support contiguous inter-domain TE LSPs since this is the default mode of operation for RSVP-TE. Failure to support either or both of nested LSPs or stitched LSPs, restricts the operators options, but does not prevent the establishment of inter-domain TE LSPs.

Signaling Options

There are three ways in which an RSVP-TE LSP could be signaled across multiple domains:

Contiguous

  A contiguous TE LSP is a single TE LSP that is set up across
  multiple domains using RSVP-TE signaling procedures described in
  RFC3209 and RFC3473.  No additional TE LSPs are required to
  create a contiguous TE LSP, and the same RSVP-TE information for
  the TE LSP is maintained along the entire LSP path.  In
  particular, the TE LSP has the same RSVP-TE session and LSP ID at
  every LSR along its path.

Nested

  One or more TE LSPs may be nested within another TE LSP as
  described in RFC4206.  This technique can be used to nest one or
  more inter-domain TE LSPs into an intra-domain hierarchical LSP
  (H-LSP).  The label stacking construct is used to achieve nesting
  in packet networks.  In the rest of this document, the term H-LSP
  is used to refer to an LSP that allows other LSPs to be nested
  within it.  An H-LSP may be advertised as a TE link within the
  same instance of the routing protocol as was used to advertise the
  TE links from which it was created, in which case it is a
  Forwarding Adjacency (FA) RFC4206.

Stitched

  The concept of LSP stitching as well as the required signaling
  procedures are described in RFC5150.  This technique can be used
  to stitch together shorter LSPs (LSP segments) to create a single,
  longer LSP.  The LSP segments of an inter-domain LSP may be
  intra-domain LSPs or inter-domain LSPs.

  The process of stitching in the data plane results in a single,
  end-to-end contiguous LSP.  But in the control plane, each segment
  is signaled as a separate LSP (with distinct RSVP sessions) and
  the end-to-end LSP is signaled as yet another LSP with its own
  RSVP session.  Thus, the control plane operation for LSP stitching
  is very similar to that for nesting.

An end-to-end inter-domain TE LSP may be achieved using one or more of the signaling techniques described. The choice is a matter of policy for the node requesting LSP setup (the ingress) and policy for each successive domain border node. On receipt of an LSP setup request (RSVP-TE Path message) for an inter-domain TE LSP, the decision of whether to signal the LSP contiguously or whether to nest or stitch it to another TE LSP depends on the parameters signaled from the ingress node and on the configuration of the local node.

The stitching segment LSP or H-LSP used to cross a domain may be pre-established or signaled dynamically based on the demand caused by the arrival of the inter-domain TE LSP setup request.

Procedures on the Domain Border Node

Whether an inter-domain TE LSP is contiguous, nested, or stitched is limited by the signaling methods supported by or configured on the intermediate nodes. It is usually the domain border nodes where this restriction applies since other transit nodes are oblivious to the mechanism in use. The ingress of the LSP may further restrict the choice by setting parameters in the Path message when it is signaled.

When a domain border node receives the RSVP Path message for an inter-domain TE LSP setup, it MUST carry out the following procedures before it can forward the Path message to the next node along the path:

  1.  Apply policies for the domain and the domain border node.
      These policies may restrict the establishment of inter-domain
      TE LSPs.  In case of a policy failure, the node SHOULD fail
      the setup and send a PathErr message with error code "Policy
      control failure"/ "Inter-domain policy failure".

  2.  Determine the signaling method to be used to cross the domain.
      If the ingress node of the inter-domain TE LSP has specified
      restrictions on the methods to be used, these MUST be adhered
      to.  Within the freedom allowed by the ingress node, the
      domain border node MAY choose any method according to local
      configuration and policies.  If no resultant signaling method
      is available or allowed, the domain border node MUST send a
      PathErr message with an error code as described in Section
      4.1.

      Thus, for example, an ingress may request a contiguous LSP
      because it wishes to exert maximal control over the LSP's path
      and to control when reoptimization takes place.  But the
      operator of a transit domain may decide (for example) that
      only LSP stitching is allowed for exactly the reason that it
      gives the operator the chance to reoptimize their own domain
      under their own control.  In this case, the policy applied at
      the entry to the transit domain will result in the return of a
      PathErr message and the ingress has a choice to:

      - find another path avoiding the transit domain,
      - relax his requirements, or
      - fail to provide the service.

  3.  Carry out ERO procedures as described in Section 3 in addition
      to the procedures in RFC3209 and RFC3473.

  4.  Perform any path computations as required to determine the
      path across the domain and potentially to select the exit
      point from the domain.

      The path computation procedure is outside the scope of this
      document.  A path computation option is specified in
      RFC5152, and another option is to use a Path Computation
      Element (PCE) RFC4655.

     4a.  In the case of nesting or stitching, either find an
          existing intra-domain TE LSP to carry the inter-domain TE
          LSP or signal a new one, depending on local policy.

      In the event of a path computation failure, a PathErr message
      SHOULD be sent with error code "Routing Problem" using an
      error value selected according to the reason for computation
      failure.  A domain border node MAY opt to silently discard the
      Path message in this case as described in Section 8.

In the event of the receipt of a PathErr message reporting signaling failure from within the domain or reported from a downstream domain, the domain border node MAY apply crankback procedures as described in Section 3.2. If crankback is not applied, or is exhausted, the border node MUST continue with PathErr processing as described in RFC3209 and RFC3473.

In the event of successful processing of a Path or Resv message, the domain border node MUST carry out RRO procedures as described in Section 3.3.

Rules on ERO Processing

The ERO that a domain border node receives in the Path message was supplied by the ingress node of the TE LSP and may have been updated by other nodes (for example, other domain border nodes) as the Path message was propagated. The content of the ERO depends on several factors including:

- the path computation techniques used, - the degree of TE visibility available to the nodes performing path

 computation, and

- the policy at the nodes creating/modifying the ERO.

In general, H-LSPs and LSP segments are used between domain border nodes, but there is no restriction on the use of such LSPs to span multiple hops entirely within a domain. Therefore, the discussion that follows may be equally applied to any node within a domain although the term "domain border node" continues to be used for clarity.

When a Path message reaches the domain border node, the following rules apply for ERO processing and for further signaling.

  1.  If there are any policies related to ERO processing for the
      LSP, they MUST be applied and corresponding actions MUST be
      taken.  For example, there might be a policy to reject EROs
      that identify nodes within the domain.  In case of
      inter-domain LSP setup failures due to policy failures related
      to ERO processing, the node SHOULD issue a PathErr with error
      code "Policy control failure"/"Inter-domain explicit route
      rejected", but MAY be configured to silently discard the Path
      message or to return a different error code for security
      reasons.

  2.  Section 8.2 of RFC4206 describes how a node at the edge of a
      region processes the ERO in the incoming Path message and uses
      this ERO, to either find an existing H-LSP or signal a new
      H-LSP using the ERO hops.  This process includes adjusting the
      ERO before sending the Path message to the next hop.  These
      procedures MUST be followed for nesting or stitching of
      inter-domain TE LSPs.

  3.  If an ERO subobject identifies a TE link formed by the
      advertisement of an H-LSP or LSP segment (whether numbered or
      unnumbered), contiguous signaling MUST NOT be used.  The node
      MUST use either nesting or stitching according to the
      capabilities of the LSP that forms the TE link, the parameters
      signaled in the Path message, and local policy.  If there is a
      conflict between the capabilities of the LSP that forms the TE
      link indicated in the ERO and the parameters on the Path
      message, the domain border node SHOULD send a PathErr with
      error code "Routing Problem"/"ERO conflicts with inter-domain
      signaling method", but MAY be configured to silently discard
      the Path message or to return a different error code for
      security reasons.

  4.  An ERO in a Path message received by a domain border node may
      have a loose hop as the next hop.  This may be an IP address
      or an AS number.  In such cases, the ERO MUST be expanded to
      determine the path to the next hop using some form of path
      computation that may, itself, generate loose hops.

  5.  In the absence of any ERO subobjects beyond the local domain
      border node, the LSP egress (the destination encoded in the
      RSVP Session object) MUST be considered as the next loose hop
      and rule 4 applied.

  6.  In the event of any other failures processing the ERO, a
      PathErr message SHOULD be sent as described in RFC3209 or
      RFC3473, but a domain border router MAY be configured to
      silently discard the Path message or to return a different
      error code for security reasons.

LSP Setup Failure and Crankback

When an error occurs during LSP setup, a PathErr message is sent back towards the LSP ingress node to report the problem. If the LSP traverses multiple domains, this PathErr will be seen successively by each domain border node.

Domain border nodes MAY apply local policies to restrict the propagation of information about the contents of the domain. For example, a domain border node MAY replace the information in a PathErr message that indicates a specific failure at a specific node with information that reports a more general error with the entire domain. These procedures are similar to those described for the borders of overlay networks in RFC4208.

However:

- A domain border node MUST NOT suppress the propagation of a PathErr

 message except to attempt rerouting as described below.

- Nodes other than domain border nodes SHOULD NOT modify the contents

 of a PathErr message.

- Domain border nodes SHOULD NOT modify the contents of a PathErr

 message unless domain confidentiality is a specific requirement.

Domain border nodes provide an opportunity for crankback rerouting RFC4920. On receipt of a PathErr message generated because of an LSP setup failure, a domain border node MAY hold the PathErr and make further attempts to establish the LSP if allowed by local policy and by the parameters signaled on the Path message for the LSP. Such attempts might involve the computation of alternate routes through the domain, or the selection of different downstream domains. If a subsequent attempt is successful, the domain border router MUST discard the held PathErr message, but if all subsequent attempts are unsuccessful, the domain border router MUST send the PathErr upstream toward the ingress node. In this latter case, the domain border router MAY change the information in the PathErr message to provide further crankback details and information aggregation as described in RFC4920.

Crankback rerouting MAY also be used to handle the failure of LSPs after they have been established RFC4920.

RRO Processing across Domains

RFC3209 defines the RRO as an optional object used for loop detection and for providing information about the hops traversed by LSPs.

As described for overlay networks in RFC4208, a domain border node MAY filter or modify the information provided in an RRO for confidentiality reasons according to local policy. For example, a series of identifiers of hops within a domain MAY be replaced with the domain identifier (such as the AS number) or be removed entirely leaving just the domain border nodes.

Note that a domain border router MUST NOT mask its own presence, and MUST include itself in the RRO.

Such filtering of RRO information does not hamper the working of the signaling protocol, but the subsequent information loss may render management diagnostic procedures inoperable or at least make them more complicated, requiring the coordination of administrators of multiple domains.

Similarly, protocol procedures that depend on the presence of RRO information may become inefficient. For example, the Fast Reroute procedures defined in RFC4090 use information in the RRO to determine the labels to use and the downstream MP.

Notify Message Processing

Notify messages are introduced in RFC3473. They may be sent direct rather than hop-by-hop, and so may speed the propagation of error information. If a domain border router is interested in seeing such messages (for example, to enable it to provide protection switching), it is RECOMMENDED that the domain border router update the Notify Request objects in the Path and Resv messages to show its own address following the procedures of RFC3473.

Note that the replacement of a Notify Recipient in the Notify Request object means that some Notify messages (for example, those intended for delivery to the ingress LSR) may need to be examined, processed, and forwarded at domain borders. This is an obvious trade-off issue as the ability to handle notifiable events locally (i.e., within the domain) may or may not outweigh the cost of processing and forwarding Notify messages beyond the domain. Observe that the cost increases linearly with the number of domains in use.

Also note that, as described in Section 8, a domain administrator may wish to filter or modify Notify messages that are generated within a domain in order to preserve security or confidentiality of network information. This is most easily achieved if the Notify messages are sent via the domain borders.

RSVP-TE Signaling Extensions

The following RSVP-TE signaling extensions are defined to enable inter-domain LSP setup.

Control of Choice of Signaling Method

In many network environments, there may be a network-wide policy that determines which one of the three inter-domain LSP techniques is used. In these cases, no protocol extensions are required.

However, in environments that support more than one technique, an ingress node may wish to constrain the choice made by domain border nodes for each inter-domain TE LSP that it originates.

RFC4420 defines the LSP_Attributes object that can be used to signal required attributes of an LSP. The Attributes Flags TLV includes Boolean flags that define individual attributes.

This document defines a new bit in the TLV that can be set by the ingress node of an inter-domain TE LSP to restrict the intermediate nodes to using contiguous signaling:

  Contiguous LSP bit (bit number assignment in Section 9.1)

This flag is set by the ingress node that originates a Path message to set up an inter-domain TE LSP if it requires that the contiguous LSP technique is used. This flag bit is only to be used in the Attributes Flags TLV.

When a domain border LSR receives a Path message containing this bit set (one), the node MUST NOT perform stitching or nesting in support of the inter-domain TE LSP being set up. When this bit is clear (zero), a domain border LSR MAY perform stitching or nesting according to local policy.

This bit MUST NOT be modified by any transit node.

An intermediate node that supports the LSP_Attributes object and the Attributes Flags TLV, and also recognizes the "Contiguous LSP" bit, but cannot support contiguous TE LSPs, MUST send a Path Error message with an error code "Routing Problem"/"Contiguous LSP type not supported" if it receives a Path message with this bit set.

If an intermediate node receiving a Path message with the "Contiguous LSP" bit set in the Flags field of the LSP_Attributes, recognizes the object, the TLV, and the bit and also supports the desired contiguous LSP behavior, then it MUST signal a contiguous LSP. If the node is a domain border node, or if the node expands a loose hop in the ERO, it MUST include an RRO Attributes subobject in the RRO of the corresponding Resv message (if such an object is present) with the "Contiguous LSP" bit set to report its behavior.

Domain border LSRs MUST support and act on the setting of the "Contiguous LSP" flag.

However, if the intermediate node supports the LSP_Attributes object but does not recognize the Attributes Flags TLV, or supports the TLV but does not recognize this "Contiguous LSP" bit, then it MUST forward the object unmodified.

The choice of action by an ingress node that receives a PathErr when requesting the use of a contiguous LSP is out of the scope of this document, but may include the computation of an alternate path.

Protection and Recovery of Inter-Domain TE LSPs

The procedures described in Sections 3 and 4 MUST be applied to all inter-domain TE LSPs, including bypass tunnels, detour LSPs RFC4090, and segment recovery LSPs RFC4873. This means that these LSPs will also be subjected to ERO processing, policies, path computation, etc.

Note also that the paths for these backup LSPs need to be either pre-configured, computed, and signaled with the protected LSP or computed on-demand at the PLR. Just as with any inter-domain TE LSP, the ERO may comprise strict or loose hops and will depend on the TE visibility of the computation point into the subsequent domain.

If loose hops are present in the path of the backup LSP, ERO expansion will be required at some point along the path: probably at a domain border node. In order that the backup path remains disjoint from the protected LSP(s) the node performing the ERO expansion must

be provided with the path of the protected LSPs between the PLR and the MP. This information can be gathered from the RROs of the protected LSPs and is signaled in the DETOUR object for Fast Reroute RFC4090 and uses route exclusion RFC4874 for other protection schemes.

Fast Recovery Support Using MPLS-TE Fast Reroute (FRR)

RFC4090 describes two methods for local protection for a packet TE LSP in case of link, Shared Risk Link Group (SRLG), or node failure. This section describes how these mechanisms work with the proposed signaling solutions for inter-domain TE LSP setup.

Failure within a Domain (Link or Node Failure)

The mode of operation of MPLS-TE Fast Reroute to protect a contiguous, stitched, or nested TE LSP within a domain is identical to the existing procedures described in RFC4090. Note that, in the case of nesting or stitching, the end-to-end LSP is automatically protected by the protection operation performed on the H-LSP or stitching segment LSP.

No protocol extensions are required.

Failure of a Link at a Domain Border

This case arises where two domains are connected by a TE link. In this case, each domain has its own domain border node, and these two nodes are connected by the TE link. An example of this case is where the ASBRs of two ASs are connected by a TE link.

A contiguous LSP can be backed up using any PLR and MP, but if the LSP uses stitching or nesting in either of the connected domains, the PLR and MP MUST be domain border nodes for those domains. It will be usual to attempt to use the local (connected by the failed link) domain border nodes as the PLR and MP.

To protect an inter-domain link with MPLS-TE Fast Reroute, a set of backup tunnels must be configured or dynamically computed between the PLR and MP such that they are diversely routed from the protected inter-domain link and the protected inter-domain LSPs.

Each protected inter-domain LSP using the protected inter-domain TE link must be assigned to an NHOP bypass tunnel that is diverse from the protected LSP. Such an NHOP bypass tunnel can be selected by analyzing the RROs in the Resv messages of the available bypass

tunnels and the protected TE LSP. It may be helpful to this process if the extensions defined in RFC4561 are used to clearly distinguish nodes and links in the RROs.

Failure of a Border Node

Two border node failure cases exist. If the domain border falls on a link as described in the previous section, the border node at either end of the link may fail. Alternatively, if the border falls on a border node (as is the case with IGP areas), that single border node may fail.

It can be seen that if stitching or nesting is used, the failed node will be the start or end (or both) of a stitching segment LSP or H-LSP, in which case protection must be provided to the far end of the stitching segment or H-LSP. Thus, where one of these two techniques is in use, the PLR will be the upstream domain entry point in the case of the failure of the domain exit point, and the MP will be the downstream domain exit point in the case of the failure of the domain entry point. Where the domain border falls at a single domain border node, both cases will apply.

If the contiguous LSP mechanism is in use, normal selection of the PLR and MP can be applied, and any node within the domains may be used to fill these roles.

As before, selection of a suitable backup tunnel (in this case, an NNHOP backup) must consider the paths of the backed-up LSPs and the available NNHOP tunnels by examination of their RROs.

Note that where the PLR is not immediately upstream of the failed node, error propagation time may be delayed unless some mechanism such as [BFD-MPLS] is implemented or unless direct reporting, such as through the GMPLS Notify message RFC3473, is employed.

Protection and Recovery of GMPLS LSPs

RFC4873 describes GMPLS-based segment recovery. This allows protection against a span failure, a node failure, or failure over any particular portion of a network used by an LSP.

The domain border failure cases described in Section 5.1 may also occur in GMPLS networks (including packet networks) and can be protected against using segment protection without any additional protocol extensions.

Note that if loose hops are used in the construction of the working and protection paths signaled for segment protection, then care is required to keep these paths disjoint. If the paths are signaled incrementally, then route exclusion RFC4874 may be used to ensure that the paths are disjoint. Otherwise, a coordinated path computation technique such as that offered by cooperating Path Computation Elements RFC4655 can provide suitable paths.

Reoptimization of Inter-Domain TE LSPs

Reoptimization of a TE LSP is the process of moving the LSP from the current path to a more preferred path. This involves the determination of the preferred path and make-before-break signaling procedures RFC3209 to minimize traffic disruption.

Reoptimization of an inter-domain TE LSP may require a new path in more than one domain.

The nature of the inter-domain LSP setup mechanism defines how reoptimization can be applied. If the LSP is contiguous, then the signaling of the make-before-break process MUST be initiated by the ingress node as defined in RFC3209. But if the reoptimization is limited to a change in path within one domain (that is, if there is no change to the domain border nodes) and nesting or stitching is in use, the H-LSP or stitching segment may be independently reoptimized within the domain without impacting the end-to-end LSP.

In all cases, however, the ingress LSR may wish to exert control and coordination over the reoptimization process. For example, a transit domain may be aware of the potential for reoptimization, but not bother because it is not worried by the level of service being provided across the domain. But the cumulative effect on the end-to-end LSP may cause the head-end to worry and trigger an end-to-end reoptimization request (of course, the transit domain may choose to ignore the request).

Another benefit of end-to-end reoptimization over per-domain reoptimization for non-contiguous inter-domain LSPs is that per-domain reoptimization is restricted to preserve the domain entry and exit points (since to do otherwise would break the LSP!). But end-to-end reoptimization is more flexible and can select new domain border LSRs.

There may be different cost-benefit analysis considerations between end-to-end reoptimization and per-domain reoptimization. The greater the number of hops involved in the reoptimization, the higher the risk of traffic disruption. The shorter the segment reoptimized, the lower the chance of making a substantial improvement on the quality of the end-to-end LSP. Administrative policies should be applied in this area with care.

RFC4736 describes mechanisms that allow:

- The ingress node to request each node with a loose next hop to

 re-evaluate the current path in order to search for a more optimal
 path.

- A node with a loose next hop to inform the ingress node that a

 better path exists.

These mechanisms SHOULD be used for reoptimization of a contiguous inter-domain TE LSP.

Note that end-to-end reoptimization may involve a non-local modification that might select new entry / exit points. In this case, we can observe that local reoptimization is more easily and flexibly achieved using nesting or stitching. Further, the "locality principle" (i.e., the idea of keeping information only where it is needed) is best achieved using stitching or nesting. That said, a contiguous LSP can easily be modified to take advantage of local reoptimizations (as defined in RFC4736) even if this would require the dissemination of information and the invocation of signaling outside the local domain.

Backward Compatibility

The procedures in this document are backward compatible with existing deployments.

- Ingress LSRs are not required to support the extensions in this

 document to provision intra-domain LSPs.  The default behavior by
 transit LSRs that receive a Path message that does not have the
 "Contiguous LSP" bit set in the Attributes Flags TLV of the
 LSP_Attributes object or does not even have the object present is
 to allow all modes of inter-domain TE LSP, so back-level ingress
 LSRs are able to initiate inter-domain LSPs.

- Transit, non-border LSRs are not required to perform any special

 processing and will pass the LSP_Attributes object onwards
 unmodified according to the rules of RFC2205.  Thus, back-level
 transit LSRs are fully supported.

- Domain border LSRs will need to be upgraded before inter-domain TE

 LSPs are allowed.  This is because of the need to establish policy,
 administrative, and security controls before permitting
 inter-domain LSPs to be signaled across a domain border.  Thus,
 legacy domain border LSRs do not need to be considered.

The RRO additions in this document are fully backward compatible.

Security Considerations

RSVP does not currently provide for automated key management. RFC4107 states a requirement for mandatory automated key management under certain situations. There is work starting in the IETF to define improved authentication including automated key management for RSVP. Implementations and deployments of RSVP should pay attention to any capabilities and requirements that are outputs from this ongoing work.

A separate document is being prepared to examine the security aspects of RSVP-TE signaling with special reference to multi-domain scenarios [MPLS-SEC]. RFC4726 provides an overview of the requirements for security in an MPLS-TE or GMPLS multi-domain environment.

Before electing to utilize inter-domain signaling for MPLS-TE, the administrators of neighboring domains MUST satisfy themselves as to the existence of a suitable trust relationship between the domains. In the absence of such a relationship, the administrators SHOULD decide not to deploy inter-domain signaling, and SHOULD disable RSVP-TE on any inter-domain interfaces.

When signaling an inter-domain RSVP-TE LSP, an operator MAY make use of the security features already defined for RSVP-TE RFC3209. This may require some coordination between the domains to share the keys (see RFC2747 and RFC3097), and care is required to ensure that the keys are changed sufficiently frequently. Note that this may involve additional synchronization, should the domain border nodes be protected with FRR, since the MP and PLR should also share the key.

For an inter-domain TE LSP, especially when it traverses different administrative or trust domains, the following mechanisms SHOULD be provided to an operator (also see RFC4216):

1) A way to enforce policies and filters at the domain borders to

  process the incoming inter-domain TE LSP setup requests (Path
  messages) based on certain agreed trust and service
  levels/contracts between domains.  Various LSP attributes such as
  bandwidth, priority, etc. could be part of such a contract.

2) A way for the operator to rate-limit LSP setup requests or error

  notifications from a particular domain.

3) A mechanism to allow policy-based outbound RSVP message processing

  at the domain border node, which may involve filtering or
  modification of certain addresses in RSVP objects and messages.

Additionally, an operator may wish to reduce the signaling interactions between domains to improve security. For example, the operator might not trust the neighboring domain to supply correct or trustable restart information RFC5063 and might ensure that the availability of restart function is not configured in the Hello message exchange across the domain border. Thus, suitable configuration MUST be provided in an RSVP-TE implementation to enable the operator to control optional protocol features that may be considered a security risk.

Some examples of the policies described above are as follows:

 A) An operator may choose to implement some kind of ERO filtering
    policy on the domain border node to disallow or ignore hops
    within the domain from being identified in the ERO of an
    incoming Path message.  That is, the policy is that a node
    outside the domain cannot specify the path of the LSP inside the
    domain.  The domain border LSR can make implement this policy in
    one of two ways:

      - It can reject the Path message.

      - It can ignore the hops in the ERO that lie within the
        domain.

 B) In order to preserve confidentiality of network topology, an
    operator may choose to disallow recording of hops within the
    domain in the RRO or may choose to filter out certain recorded
    RRO addresses at the domain border node.

 C) An operator may require the border node to modify the addresses
    of certain messages like PathErr or Notify originated from hops
    within the domain.

 D) In the event of a path computation failure, an operator may
    require the border node to silently discard the Path message
    instead of returning a PathErr.  This is because a Path message
    could be interpreted as a network probe, and a PathErr provides
    information about the network capabilities and policies.

Note that the detailed specification of such policies and their implementation are outside the scope of this document.

Operations, Administration, and Management (OAM) mechanisms including [BFD-MPLS] and RFC4379 are commonly used to verify the connectivity of end-to-end LSPs and to trace their paths. Where the LSPs are inter-domain LSPs, such OAM techniques MAY require that OAM messages are intercepted or modified at domain borders, or are passed transparently across domains. Further discussion of this topic can be found in [INTERAS-PING] and [MPLS-SEC].

IANA Considerations

IANA has made the codepoint allocations described in the following sections.

Attribute Flags for LSP_Attributes Object

A new bit has been allocated from the "Attributes Flags" sub-registry of the "RSVP TE Parameters" registry.

 Bit | Name                 | Attribute  | Path       | RRO | Reference
 No  |                      | Flags Path | Flags Resv |     |
 ----+----------------------+------------+------------+-----+----------
 4     Contiguous LSP         Yes          No           Yes   RFC5150

New Error Codes

New RSVP error codes/values have been allocated from the "Error Codes and Globally-Defined Error Value Sub-Codes" sub-registry of the "RSVP Parameters" registry.

For the existing error code "Policy control failure" (value 2), two new error values have been registered as follows:

  103 = Inter-domain policy failure
  104 = Inter-domain explicit route rejected

For the existing error code "Routing Problem" (value 24), two new error values have been registered as follows:

  28 = Contiguous LSP type not supported
  29 = ERO conflicts with inter-domain signaling method

10. Acknowledgements

The authors would like to acknowledge the input and helpful comments from Kireeti Kompella on various aspects discussed in the document. Deborah Brungard and Dimitri Papdimitriou provided thorough reviews.

Thanks to Sam Hartman for detailed discussions of the security considerations.

11. References

11.1. Normative References

RFC2119 Bradner, S., "Key words for use in RFCs to Indicate

               Requirement Levels", BCP 14, RFC 2119, March 1997.

RFC2205 Braden, R., Ed., Zhang, L., Berson, S., Herzog, S.,

               and S. Jamin, "Resource ReSerVation Protocol (RSVP)
               -- Version 1 Functional Specification", RFC 2205,
               September 1997.

RFC3209 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.

RFC3473 Berger, L., Ed., "Generalized Multi-Protocol Label

               Switching (GMPLS) Signaling Resource ReserVation
               Protocol-Traffic Engineering (RSVP-TE) Extensions",
               RFC 3473, January 2003.

RFC4206 Kompella, K. and Y. Rekhter, "Label Switched Paths

               (LSP) Hierarchy with Generalized Multi-Protocol Label
               Switching (GMPLS) Traffic Engineering (TE)", RFC
               4206, October 2005.

RFC4420 Farrel, A., Ed., Papadimitriou, D., Vasseur, J.-P.,

               and A. Ayyangar, "Encoding of Attributes for
               Multiprotocol Label Switching (MPLS) Label Switched
               Path (LSP) Establishment Using Resource ReserVation
               Protocol-Traffic Engineering (RSVP-TE)", RFC 4420,
               February 2006.

RFC5150 Ayyangar, A., Kompella, K., and JP. Vasseur, "Label

               Switched Path Stitching with Generalized
               Multiprotocol Label Switching Traffic Engineering
               (GMPLS TE)", RFC 5150, February 2008.

11.2. Informative References

RFC2747 Baker, F., Lindell, B., and M. Talwar, "RSVP

               Cryptographic Authentication", RFC 2747, January
               2000.

RFC3097 Braden, R. and L. Zhang, "RSVP Cryptographic

               Authentication -- Updated Message Type Value", RFC
               3097, April 2001.

RFC4090 Pan, P., Ed., Swallow, G., Ed., and A. Atlas, Ed.,

               "Fast Reroute Extensions to RSVP-TE for LSP Tunnels",
               RFC 4090, May 2005.

RFC4105 Le Roux, J.-L., Ed., Vasseur, J.-P., Ed., and J.

               Boyle, Ed., "Requirements for Inter-Area MPLS Traffic
               Engineering", RFC 4105, June 2005.

RFC4107 Bellovin, S. and R. Housley, "Guidelines for

               Cryptographic Key Management", BCP 107, RFC 4107,
               June 2005.

RFC4208 Swallow, G., Drake, J., Ishimatsu, H., and Y.

               Rekhter, "Generalized Multiprotocol Label Switching
               (GMPLS) User-Network Interface (UNI): Resource
               ReserVation Protocol-Traffic Engineering (RSVP-TE)
               Support for the Overlay Model", RFC 4208, October
               2005.

RFC4216 Zhang, R., Ed., and J.-P. Vasseur, Ed., "MPLS Inter-

               Autonomous System (AS) Traffic Engineering (TE)
               Requirements", RFC 4216, November 2005.

RFC4379 Kompella, K. and G. Swallow, "Detecting Multi-

               Protocol Label Switched (MPLS) Data Plane Failures",
               RFC 4379, February 2006.

RFC4561 Vasseur, J.-P., Ed., Ali, Z., and S. Sivabalan,

               "Definition of a Record Route Object (RRO) Node-Id
               Sub-Object", RFC 4561, June 2006.

RFC4655 Farrel, A., Vasseur, J.-P., and J. Ash, "A Path

               Computation Element (PCE)-Based Architecture", RFC
               4655, August 2006.

RFC4726 Farrel, A., Vasseur, J.-P., and A. Ayyangar, "A

               Framework for Inter-Domain Multiprotocol Label
               Switching Traffic Engineering", RFC 4726, November
               2006.

RFC4736 Vasseur, JP., Ed., Ikejiri, Y., and R. Zhang,

               "Reoptimization of Multiprotocol Label Switching
               (MPLS) Traffic Engineering (TE) Loosely Routed Label
               Switched Path (LSP)", RFC 4736, November 2006.

RFC4873 Berger, L., Bryskin, I., Papadimitriou, D., and A.

               Farrel, "GMPLS Segment Recovery", RFC 4873, May 2007.

RFC4874 Lee, CY., Farrel, A., and S. De Cnodder, "Exclude

               Routes - Extension to Resource ReserVation Protocol-
               Traffic Engineering (RSVP-TE)", RFC 4874, April 2007.

RFC4920 Farrel, A., Ed., Satyanarayana, A., Iwata, A.,

               Fujita, N., and G. Ash, "Crankback Signaling
               Extensions for MPLS and GMPLS RSVP-TE", RFC 4920,
               July 2007.

[BFD-MPLS] Aggarwal, R., Kompella, K., Nadeau, T., and G.

               Swallow, "BFD For MPLS LSPs", Work in Progress,
               February 2005.

[INTERAS-PING] Nadeau, T. and G. Swallow, "Detecting MPLS Data Plane

               Failures in Inter-AS and inter-provider Scenarios",
               Work in Progress, October 2006.

RFC5152 Vasseur, JP., Ed., Ayyangar, A., Ed., and R. Zhang,

               "A Per-Domain Path Computation Method for
               Establishing Inter-Domain Traffic Engineering (TE)
               Label Switched Paths (LSPs)", RFC 5152, February
               2008.

[MPLS-SEC] Fang, L., Ed., Behringer, M., Callon, R., Le Roux, J.

               L., Zhang, R., Knight, P., Stein, Y., Bitar, N., and
               R. Graveman., "Security Framework for MPLS and GMPLS
               Networks", Work in Progress, July 2007.

RFC5063 Satyanarayana, A., Ed., and R. Rahman, Ed.,

               "Extensions to GMPLS Resource Reservation Protocol
               (RSVP) Graceful Restart", RFC 5063, October 2007.

Authors' Addresses

Adrian Farrel Old Dog Consulting

EMail: [email protected]

Arthi Ayyangar Juniper Networks 1194 N. Mathilda Avenue Sunnyvale, CA 94089 USA

EMail: [email protected]

Jean Philippe Vasseur Cisco Systems, Inc. 300 Beaver Brook Road Boxborough , MA - 01719 USA

EMail: [email protected]

Full Copyright Statement

Copyright (C) The IETF Trust (2008).

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, THE IETF TRUST AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Intellectual Property

The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79.

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

The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at [email protected].