Network Working Group T. Takeda, Ed.
Request for Comments: 5298 NTT
Category: Informational A. Farrel, Ed.
Old Dog Consulting
Cisco Systems, Inc.
August 2008 Analysis of Inter-Domain Label Switched Path (LSP) Recovery
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.
Protection and recovery are important features of service offerings
in Multiprotocol Label Switching (MPLS) and Generalized MPLS (GMPLS)
networks. Increasingly, MPLS and GMPLS networks are being extended
from single domain scope to multi-domain environments.
Various schemes and processes have been developed to establish Label
Switched Paths (LSPs) in multi-domain environments. These are
discussed in RFC 4726: "A Framework for Inter-Domain Multiprotocol
Label Switching Traffic Engineering".
This document analyzes the application of these techniques to
protection and recovery in multi-domain networks. The main focus for
this document is on establishing end-to-end diverse Traffic
Engineering (TE) LSPs in multi-domain networks.
Protection and recovery in Multiprotocol Label Switching (MPLS) and
Generalized MPLS (GMPLS) networks are described in [RFC4428]. These
are important features for service delivery in MPLS and GMPLS
MPLS and GMPLS networks were originally limited to single domain
environments. Increasingly, multi-domain MPLS and GMPLS networks are
being considered, where a domain is considered to be any collection
of network elements within a common sphere of address management or
path computational responsibility. Examples of such domains include
Interior Gateway Protocol (IGP) areas and Autonomous Systems (ASes).
[RFC4726] provides a framework for inter-domain MPLS and GMPLS
traffic engineering. It introduces and discusses the various schemes
and processes to establish Label Switched Paths (LSPs) in multi-
However, protection and recovery introduce additional complexities to
LSP establishment. Protection LSPs are generally required to be path
diverse from the working LSPs that they protect. Achieving this is
particularly challenging in multi-domain environments because no
single path computation or planning point is capable of determining
path diversity for both paths from one end to the other.
This document analyzes various schemes to realize MPLS and GMPLS LSP
recovery in multi-domain networks. The main focus for this document
is on establishing end-to-end diverse Traffic Engineering (TE) LSPs
in multi-domain networks.
The reader is assumed to be familiar with the terminology for LSP
recovery set out in [RFC4427], and with the terms introduced in
[RFC4726] that provides a framework for inter-domain Label Switched
Path (LSP) setup. Key terminology may also be found in [RFC4216]
that sets out requirements for inter-AS MPLS traffic engineering.
The following key terms from those sources are used within this
- Domain: See [RFC4726]. A domain is considered to be any collection
of network elements within a common sphere of address management or
path computational responsibility. Note that nested domains
continue to be out of scope. Section 1.2 provides additional
- Working LSP: See [RFC4427]. The working LSP transports normal user
traffic. Note that the term LSP and TE LSP will be used
- Recovery LSP: See [RFC4427]. The recovery LSP transports normal
user traffic when the working LSP fails. The recovery LSP may also
carry user traffic even when the working LSP is operating normally
and transporting the user traffic (e.g., 1+1 protection). The
recovery LSP is sometimes referred to as a protecting LSP.
- Diversity: See [RFC4726]. Diversity means the relationship of
multiple LSPs, where those LSPs do not share some specific type of
resource (e.g., link, node, or shared risk link group (SRLG)).
Diversity is also referred to as disjointness.
Diverse LSPs may be used for various purposes, such as load-
balancing and recovery. In this document, the recovery aspect of
diversity, specifically the end-to-end diversity of two traffic
engineering (TE) LSPs, is the focus. The two diverse LSPs are
referred to as the working LSP and recovery LSP.
- Confidentiality: See [RFC4216]. Confidentiality refers to the
protection of information about the topology and resources of one
domain from visibility by people or applications outside that
In order to fully understand the issues addressed in this document,
it is necessary to carefully define and scope the term "domain".
As defined in [RFC4726], a domain is considered to be any collection
of network elements within a common sphere of address management or
path computational responsibility. Examples of such domains include
IGP areas and Autonomous Systems. Networks accessed over the GMPLS
User-to-Network Interface (UNI) [RFC4208], and Layer One Virtual
Private Networks (L1VPNs) [RFC4847] are special cases of multi-domain
Example motivations for using more than one domain include
administrative separation, scalability, and the construction of
domains for the purpose of providing protection. These latter
"protection domains" offer edge-to-edge protection facilities for
spans or segments of end-to-end LSPs.
As described in [RFC4726], there could be TE parameters (such as
color and priority) whose meanings are specific to each domain. In
such scenarios, in order to set up inter-domain LSPs, mapping
functions may be needed to transform the TE parameters based on
policy agreements between domain administrators.
1.3. Document Scope
This document analyzes various schemes to realize MPLS and GMPLS LSP
recovery in multi-domain networks. It is based on the existing
framework for multi-domain LSP setup [RFC4726]. Note that this
document does not prevent the development of additional techniques
where appropriate (i.e., additional to the ones described in this
document). In other words, this document shows how the existing
techniques can be applied.
There are various recovery techniques for LSPs. For TE LSPs, the
major techniques are end-to-end recovery [RFC4872], local protection
such as Fast Reroute (FRR) [RFC4090] (in packet switching
environments), and segment recovery [RFC4873] (in GMPLS).
The main focus of this document is the analysis of diverse TE LSP
setup schemes that can be used in the context of end-to-end recovery.
The methodology is to show different combinations of functional
elements such as path computation and signaling techniques.
[RFC4105] and [RFC4216] describe requirements for diverse LSPs.
There are various types of diversity, and this document focuses on
node, link, and shared risk link group (SRLG) diversity.
Recovery LSPs may be used for 1+1 protection, 1:1 protection, or
shared mesh restoration. However, the requirements for path
diversity, the ways to compute diverse paths, and the signaling of
these TE LSPs are common across all uses. These topics are the main
scope of this document.
Note that diverse LSPs may be used for various purposes in addition
to recovery. An example is for load-balancing, so as to limit the
traffic disruption to a portion of the traffic flow in case of a
single node failure. This document does not preclude use of diverse
LSP setup schemes for other purposes.
The following are beyond the scope of this document.
- Analysis of recovery techniques other than the use of link, node,
or SRLG diverse LSPs (see Section 1.4).
- Details of maintenance of diverse LSPs, such as re-optimization and
Operations and Maintenance (OAM).
- Comparative evaluation of LSP setup schemes.
1.4. Note on Other Recovery Techniques
Fast Reroute and segment recovery in multi-domain networks are
described in Section 5.4 of [RFC4726], and a more detailed analysis
is provided in Section 5 of [RFC5151]. This document does not cover
any additional analysis for Fast Reroute and segment recovery in
The recovery type of an LSP or service may change at a domain
boundary. That is, the recovery type could remain the same within
one domain, but might be different in the next domain or on the
connections between domains. This may be due to the capabilities of
each domain, administrative policies, or to topology limitations. An
example is where protection at the domain boundary is provided by
link protection on the inter-domain links, but where protection
within each domain is achieved through segment recovery. This
mixture of protection techniques is beyond the scope of this
Domain diversity (that is, the selection of paths that have only the
ingress and egress domains in common) may be considered as one form
of diversity in multi-domain networks, but this is beyond the scope
of this document (see Section 2.2).
1.5. Signaling Options
There are three signaling options for establishing inter-domain TE
LSPs: nesting, contiguous LSPs, and stitching [RFC4726]. The
description in this document of diverse LSP setup is agnostic in
relation to the signaling option used, unless otherwise specified.
Note that signaling option considerations for Fast Reroute and
segment recovery are described in [RFC5151].
2. Diversity in Multi-Domain Networks
This section describes some assumptions about achieving path
diversity in multi-domain networks.
In Figure 2, domains are connected in a mesh topology. There are
multiple paths between nodes A and D, and these paths do not cross
the same domains. If inter-domain connectivity forms a mesh,
domain-level routing is required (even for the unprotected case).
This is tightly coupled with the next-hop domain resolution/discovery
mechanisms used in IP networks.
In this document, we assume that domain-level routing is given via
configuration, policy, or some external mechanism, and that this is
not part of the path computation process described later in this
Domain-level routing may also allow "domain re-entry" where a path
re-enters a domain that it has previously exited (e.g., domain#X-
domain#Y-domain#X). This requires specific considerations when
confidentiality (described in Section 3.2) is required, and is beyond
the scope of this document.
Furthermore, the working LSP and the recovery LSP may or may not be
routed along the same set of domains in the same order. In this
document, we assume that the working LSP and recovery LSP follow the
same set of domains in the same order (via configuration, policy or
some external mechanism). That is, we assume that the domain mesh
topology is reduced to a linear domain topology for each pair of
working and recovery LSPs.
- There is no assumption about the multi-domain network topology.
For example, there could be more than two domain boundary nodes or
inter-domain links (links connecting a pair of domain boundary
nodes belonging to different domains).
- It is assumed that in a multi-domain topology, the sequence of
domains that the working LSP and the recovery LSP follow must be
the same and is pre-configured.
- Domain re-entry is out of scope and is not considered further.
2.2. Note on Domain Diversity
As described in Section 1.4, domain diversity means the selection of
paths that have only the ingress and egress domains in common. This
may provide enhanced resilience against failures, and is a way to
ensure path diversity for most of the path of the LSP.
In Section 2.1, we assumed that the working LSP and the recovery LSP
follow the same set of domains in the same order. Under this
assumption, domain diversity cannot be achieved. However, by
relaxing this assumption, domain diversity could be achieved, e.g.,
by either of the following schemes.
- Consider domain diversity as a special case of SRLG diversity
(i.e., assign an SRLG ID to each domain).
- Configure domain-level routing to provide domain-diverse paths
(e.g., by means of AS_PATH in BGP).
Further details of the operation of domain diversity are beyond the
scope of this document.
3. Factors to Consider
3.1. Scalability versus Optimality
As described in [RFC4726], scalability and optimality are two
conflicting objectives. Note that the meaning of optimality differs
depending on each network operation. Some examples of optimality in
the context of diverse LSPs are:
- Minimizing the sum of their cost while maintaining diversity.
- Restricting the difference of their costs (for example, so as to
minimize the delay difference after switch-over) while maintaining
By restricting TE information distribution to only within each domain
(and not across domain boundaries) as required by [RFC4105] and
[RFC4216], it may not be possible to compute an optimal path. As
such, it might not be possible to compute diverse paths, even if they
exist. However, if we assume domain-level routing is given (as
discussed in Section 2), it would be possible to compute diverse
paths using specific computation schemes, if such paths exist. This
is discussed further in Section 4.
3.2. Key Concepts
Three key concepts to classify various diverse LSP computation and
setup schemes are presented below.
o With or without confidentiality
- Without confidentiality
It is possible to specify a path across multiple domains in
signaling (by means of the Resource Reservation Protocol-TE
(RSVP-TE) Explicit Route Object (ERO)), and to obtain record of
the inter-domain path used (by means of the RSVP-TE Record Route
Object (RRO)). In this case, it is clear that one domain has
control over the path followed in another domain, and that the
path actually used in one domain is visible from within another
Examples of this configuration are multi-area networks, and some
forms of multi-AS networks (especially within the same provider).
In these cases, there is no requirement for confidentiality.
- With confidentiality
Where confidentiality of domain topology and operational policy
is required, it is not possible to specify or obtain a full path
across other domains. Partial paths may be specified and
reported using domain identifiers or the addresses of domain
border routers in the EROs and RROs.
Examples of this configuration are some forms of multi-AS
networks (especially inter-provider networks), GMPLS-UNI
networks, and L1VPNs.
o Multi-domain path computation, per-domain path computation, and
inter-domain collaborative path computation
- Multi-domain path computation
If a single network element can see the topology of all domains
along the path, it is able to compute a full end-to-end path.
Clearly, this is only possible where confidentiality is not
Such a network element might be the head-end Label Switching
Router (LSR), a Path Computation Element (PCE) [RFC4655], or a
Network Management System (NMS). This mode of path computation
is discussed in Sections 4 and 5.
- Per-domain path computation
The path of an LSP may be computed domain-by-domain as LSP
signaling progresses through the network. This scheme requires
ERO expansion in each domain to construct the next segment of the
path toward the destination. The establishment of unprotected
LSPs in this way is covered in [RFC5152].
- Inter-domain collaborative path computation
In this scheme, path computation is typically done before
signaling and uses communication between cooperating PCEs. An
example of such a scheme is Backward Recursive Path Computation
It is possible to combine multiple path computation techniques
(including using a different technique for the working and
recovery LSPs), but details are beyond the scope of this
o Sequential path computation or simultaneous path computation
- Sequential path computation
The path of the working LSP is computed without considering the
recovery LSP, and then the path of the recovery LSP is computed.
This scheme is applicable when the recovery LSP is added later as
a change to the service grade, but may also be used when both the
working and recovery LSPs are established from the start.
Using this approach, it may not be possible to find diverse paths
for the LSPs in "trap" topologies. Furthermore, such sequential
path computation approaches reduce the likelihood of selecting an
"optimal" solution with regards to a specific objective function.
- Simultaneous path computation
The path of the working LSP and the path of the recovery LSP are
computed simultaneously. In this scheme, it is possible to avoid
trap conditions and it may be more possible to achieve an optimal
Note that LSP setup, with or without confidentiality, depends on per-
domain configuration. The choice of per-domain path computation or
inter-domain collaborative path computation, and the choice between
sequential path computation or simultaneous path computation can be
determined for each individual pair of working/recovery LSPs.
The analysis of various diverse LSP setup schemes is described in
Sections 4 and 5, based on the concepts set out above.
Some other considerations, such as network topology-specific
considerations, addressing considerations, and SRLG diversity are
described in Sections 6, 7, and 8.
4. Diverse LSP Setup Schemes without Confidentiality
This section examines schemes for diverse LSP setup based on
different path computation techniques assuming that there is no
requirement for domain confidentiality. Section 5 makes a similar
examination of schemes where domain confidentiality is required.
4.1. Management Configuration
[RFC4726] describes this path computation technique where the full
explicit paths for the working and recovery LSPs are specified by a
management application at the head-end, and no further computation or
signaling considerations are needed.
4.2. Head-End Path Computation (with Multi-Domain Visibility)
Section 3.2.1 [RFC4726] describes this path computation technique.
The full explicit paths for the working and recovery LSPs are
computed at the head-end either by the head-end itself or by a PCE.
In either case, the computing entity has full TE visibility across
multiple domains and no further computation or signaling
considerations are needed.
4.3. Per-Domain Path Computation
Sections 3.2.2, 3.2.3, and 3.3 of [RFC4726] describe this path
computation technique. More detailed procedures are described in
In this scheme, the explicit paths of the working and recovery LSPs
are specified as the complete strict paths through the source domain
followed by either of the following:
- The complete list of boundary LSRs or domain identifiers (e.g.,
AS numbers) along the paths.
- The LSP destination.
Thus, in order to navigate each domain, the path must be expanded at
each domain boundary, i.e., per-domain. This path computation is
performed by the boundary LSR or by a PCE on its behalf.
There are two schemes for establishing diverse LSPs using per-domain
computation. These are described Sections 4.3.1 and 4.3.2.
4.3.1. Sequential Path Computation
As previously noted, the main issue with sequential path computation
is that diverse paths cannot be guaranteed. Where a per-domain path
computation scheme is applied, the computation of second path needs
to be aware of the path used by the first path in order that path
diversity can be attempted.
The RSVP-TE EXCLUDE_ROUTE Object (XRO) [RFC4874] can be used when the
second path is signaled to report the details of the first path. It
should be noted that the PRIMARY_PATH_ROUTE Object defined in
[RFC4872] for end-to-end protection is not intended as a path
exclusion mechanism and should not be used for this purpose.
The process for sequential path computation is as follows:
- The working LSP is established using per-domain path computation
as described in [RFC5152]. The path of the working LSP is
available at the head-end through the RSVP-TE RRO since domain
confidentiality is not required.
- The path of the recovery LSP across the first domain is computed
excluding the resources used by the working LSP within that
domain. If a PCE is used, the resources to be avoided can be
passed to the PCE using the Exclude Route Object (XRO) extensions
to the PCE Protocol [PCEP-XRO], [PCEP].
- The recovery LSP is now signaled across the first domain as
usual, but the path of the working LSP is also conveyed in an
RSVP-TE XRO. The XRO lists nodes, links and SRLGs that must be
avoided by the LSP being signaled, and its contents are copied
from the RRO of the working LSP.
- At each subsequent domain boundary, a segment of the path of the
recovery LSP can be computed across the new domain excluding the
resources indicated in the RSVP-TE XRO.
This scheme cannot guarantee to establish diverse LSPs (even if they
could exist) because the first (working) LSP is established without
consideration of the need for a diverse recovery LSP. It is possible
to modify the path of the working LSP using the crankback techniques
[RFC4920] if the setup of the recovery LSP is blocked or if some
resources are shared.
Note that, even if a solution is found, the degree of optimality of
the solution (i.e., of the set of diverse TE LSPs) might not be
4.3.2. Simultaneous Path Computation
Simultaneous path computation gives a better likelihood of finding a
pair of diverse paths as the diversity requirement forms part of the
computation process. In per-domain path computation mechanisms,
there are several aspects to consider.
Simultaneous path computation means that the paths of the working and
recovery LSPs are computed at the same time. Since we are
considering per-domain path computation, these two paths must be
computed at the boundary of each domain.
The process for simultaneous path computation is as follows:
- The ingress LSR (or a PCE) computes a pair of diverse paths
across the first domain. If a PCE is used, PCEP offers the
ability to request disjoint paths.
- The working LSP is signaled across the first domain as usual, but
must carry with it the requirement for a disjoint recovery LSP
and the information about the path already computed for the
recovery LSP across the first domain. In particular, the domain
boundary node used by the recovery LSP must be reported.
- Each domain boundary router, in turn, computes a pair of disjoint
paths across the next domain. The working LSP is signaled as
usual, and the computed path of the recovery LSP is collected in
the signaling messages.
- When the working LSP has been set up, the full path of the
recovery LSP is returned to the head-end LSR in the signaling
messages for the working LSP. This allows the head-end LSR to
signal the recovery LSP using a full path without the need for
further path computation.
Note that the signaling protocol mechanisms do not currently exist to
collect the path of the recovery LSP during the signaling of the
working LSP. Definition of protocol mechanisms are beyond the scope
of this document, but it is believed that such mechanisms would be
simple to define and implement.
Note also that the mechanism described is still not able to guarantee
the selection of diverse paths even where they exist since, when
domains are multiply interconnected, the determination of diverse
end-to-end paths may depend on the correct selection of inter-domain
links. Crankback [RFC4920] may also be used in combination with this
scheme to improve the chances of success.
Note that even if a solution is found, the degree of optimality of
the solution (i.e., set of diverse TE LSPs) might not be maximal.
4.4. Inter-Domain Collaborative Path Computation
Collaborative path computation requires the cooperation between PCEs
that are responsible for different domains. This approach is
described in Section 3.4 of [RFC4726]. Backward recursive path
computation (BRPC) [BRPC] provides a collaborative path computation
technique where the paths of LSPs are fully determined by
communication between PCEs before the LSPs are established. Two ways
to use BRPC for diverse LSPs are described in the following sections.
4.4.1. Sequential Path Computation
In sequential path computation, the path of the working LSP is
computed using BRPC as described in [BRPC]. The path of the recovery
LSP is then computed also using BRPC with the addition that the path
of the working LSP is explicitly excluded using the XRO route
exclusion techniques described in [PCEP-XRO].
In this case, the working LSP could be set up before or after the
path of the recovery LSP is computed. In the latter case, the actual
path of the working LSP as reported in the RSVP-TE RRO should be used
when computing the path of the recovery LSP.
This scheme cannot guarantee to establish diverse LSPs (even if they
exist) because the working LSP may block the recovery LSP. In such a
scenario, re-optimization of the working and recovery LSPs may be
used to achieve fully diverse paths.
4.4.2. Simultaneous Path Computation
In simultaneous path computation, the PCEs collaborate to compute the
paths of both the working and the recovery LSPs at the same time.
Since both LSPs are computed in a single pass, the LSPs can be
signaled simultaneously of sequentially according to the preference
of the head-end LSR.
Collaborative simultaneous path computation is achieved using the
Synchronization Vector (SVEC) object in the PCE Protocol [PCEP].
This object allows two computation requests to be associated and made
dependent. The coordination of multiple computation requests within
the BRPC mechanism is not described in [BRPC]. It is believed that
it is possible to specify procedures for such coordination, but the
development of new procedures is outside the scope of this document.
This scheme can guarantee to establish diverse LSPs where they are
possible, assuming that domain-level routing is pre-determined as
described in Section 2. Furthermore, the computed set of TE LSPs can
be guaranteed to be optimal with respect to some objective functions.
5. Diverse LSP Setup Schemes with Confidentiality
In the context of this section, the term confidentiality applies to
the protection of information about the topology and resources
present within one domain from visibility by people or applications
outside that domain. This includes, but is not limited to, recording
of LSP routes, and the advertisements of TE information. The
confidentiality does not apply to the protection of user traffic.
Diverse LSP setup schemes with confidentiality are similar to ones
without confidentiality. However, several additional mechanisms are
needed to preserve confidentiality. Examples of such mechanisms are:
- Path key: A path key is used in place of a segment of the path of
an LSP when the LSP is signaled, when the path of the LSP is
reported by signaling, or when the LSP's path is generated by a
PCE. This allows the exact path of the LSP to remain
confidential through the substitution of "confidential path
segments" (CPSs) by these path keys.
[PCE-PATH-KEY] describes how path keys can be used by PCEs to
preserve path confidentiality, and [RSVP-PATH-KEY] explains how
path keys are used in signaling. Note that if path keys are
signaled in RSVP-TE EROs they must be expanded so that the
signaling can proceed. This expansion normally takes place when
the first node in the CPS is reached. The expansion of the path
key would normally be carried out by the PCE that generated the
key, and for that reason, the path key contains an identifier of
the PCE (the PCE-ID).
- LSP segment: LSP segments can be pre-established across domains
according to some management policy. The LSP segments can be
used to support end-to-end LSPs as hierarchical LSPs [RFC4206] or
as LSP stitching segments [RFC5150].
The end-to-end LSPs are signaled indicating just the series of
domains or domain border routers. Upon entry to each domain, an
existing trans-domain LSP is selected and used to carry the end-
to-end LSP across the domain.
Note that although the LSP segments are described as being pre-
established, they could be set up on demand on receipt of the
request for the end-to-end LSP at the domain border.
It is also worth noting that in schemes that result in a single
contiguous end-to-end LSP (without LSP tunneling or stitching),
the same concept of LSP segments may apply provided that ERO
expansion is applied at domain boundaries and that the path of
the LSP is not reported in the RSVP-TE RRO.
These techniques may be applied directly or may require protocol
extensions depending on the specific diverse LSP setup schemes
described below. Note that in the schemes below, the paths of the
working and recovery LSPs are not impacted by the confidentiality
5.1. Management Configuration
Although management systems may exist that can determine end-to-end
paths even in the presence of domain confidentiality, the full paths
cannot be provided to the head-end LSR for LSP signaling as this
would break the confidentiality requirements.
Thus, for LSPs that are configured through management applications,
the end-to-end path must either be constructed using LSP segments
that cross the domains, or communicated to the head-end LSR with the
use of path keys.
5.2. Head-End Path Computation (with Multi-Domain Visibility)
It is not possible for the head-end LSR to compute the full end-to-
end path of an inter-domain LSP when domain confidentiality is in use
because the LSR will not have any TE information about the other
5.3. Per-Domain Path Computation
Per-domain path computation for working and recovery LSPs is
practical with domain confidentiality. As when there are no
confidentiality restrictions, we can separate the cases of sequential
and simultaneous path computation.
5.3.1. Sequential Path Computation
In sequential path computation, we can assume that the working LSP
has its path computed and is set up using the normal per-domain
technique as described in [RFC5152]. However, because of
confidentiality issues, the full path of the working LSP is not
returned in the signaling messages and is not available to the head-
To compute a disjoint path for the recovery LSP, each domain border
node needs to know the path of the working LSP within the domain to
which it provides entry. This is easy for the ingress LSR as it has
access to the RSVP-TE RRO within first domain. In subsequent
domains, the process requires that the path of the working LSP is
somehow made available to the domain border router as the recovery
LSP is signaled. Note that the working and recovery LSPs do not use
the same border routers if the LSPs are node or SRLG diverse.
There are several possible mechanisms to achieve this.
- Path keys could be used in the RRO for the working LSP. As the
signaling messages are propagated back towards the head-end LSR,
each domain border router substitutes a path key for the segment
of the working LSP's path within its domain. Thus, the RRO
received at the head-end LSR consists of the path within the
initial domain followed by a series of path keys.
When the recovery LSP is signaled, the path keys can be included
in the ERO as exclusions. Each domain border router in turn can
expand the path key for its domain and know which resources must
be avoided. PCEP provides a protocol that can be used to request
the expansion of the path key from the domain border router used
by the working LSP, or from some third party such as a PCE.
- Instead of using path keys, each confidential path segment in the
RRO of the working LSP could be encrypted by the domain border
routers. These encrypted segments would appear as exclusions in
the ERO for the recovery LSP and could be decrypted by the domain
No mechanism currently exists in RSVP-TE for this function, which
would probably assume a domain-wide encryption key.
- The identity of the working LSP could be included in the XRO of
the recovery LSP to indicate that a disjoint path must be found.
This option could require a simple extension to the current XRO
specification [RFC4874] to allow LSP identifiers to be included.
It could also use the Association Object [RFC4872] to identify
the working LSP.
This scheme would also need a way for a domain border router to
find the path of an LSP within its domain. An efficient way to
achieve this would be to also include the domain border router
used by the working LSP in the signaling for the recovery LSP,
but other schemes based on management applications or stateful
PCEs might also be developed.
Clearly, the details of this alternative have not been specified.
5.3.2. Simultaneous Path Computation
In per-domain simultaneous path computation the path of the recovery
LSP is computed at the same time as the working LSP (i.e., as the
working LSP is signaled). The computed path of the recovery LSP is
propagated to the head-end LSR as part of the signaling process for
the working LSP, but confidentiality must be maintained, so the full
path cannot be returned. There are two options as follows.
- LSP segment: As the signaling of the working LSP progresses and
the path of the recovery LSP is computed domain-by-domain,
trans-domain LSPs can be set up for use by the recovery LSP.
When the recovery LSP is signaled, it will pick up these LSP
segments and use them for tunneling or stitching.
This mechanism needs coordination through the management plane
between domain border routers so that a router on the working
path can request the establishment of an LSP segment for use by
the protection path. This could be achieved through the TE MIB
modules [RFC3812], [RFC4802].
Furthermore, when the recovery LSP is signaled it needs to be
sure to pick up the correct LSP segment. Therefore, some form of
LSP segment identifier needs to be reported in the signaling of
the working LSP and propagated in the signaling of the recovery
LSP. Mechanisms for this do not currently exist, but would be
relatively simple to construct.
Alternatively, the LSP segments could be marked as providing
protection for the working LSP. In this case, the recovery LSP
can be signaled with the identifier of the working LSP using the
Association Object [RFC4872] enabling the correct LSP segments to
- Path key: The path of the recovery LSP can be determined and
returned to the head-end LSR just described in Section 4.4.2, but
each CPS is replaced by a path key. As the recovery path is
signaled each path key is expanded, domain-by-domain to achieve
the correct path. This requires that the entity that computes
the path of the recovery LSP (domain border LSR or PCE) is
5.4 Inter-Domain Collaborative Path Computation
Cooperative collaboration between PCEs is also applicable when domain
confidentiality is required.
5.4.1. Sequential Path Computation
In sequential cooperative path computation, the path of the working
LSP is determined first using a mechanism such as BRPC. Since domain
confidentiality is in use, the path returned may contain path keys.
When the path of the recovery LSP is computed (which may be before or
after the working LSP is signaled) the path exclusions supplied to
the PCE and exchanged between PCEs must use path keys as described in
5.4.2. Simultaneous Path Computation
As described in Section 4.4.2, diverse path computation can be
requested using the PCEP SVEC Object [PCEP], and BRPC could be
extended for inter-domain diverse path computation. However, to
conform to domain confidentiality requirements, path keys must be
used in the paths returned by the PCEs and signaled by RSVP-TE.
Note that the LSP segment approach may not be applicable here because
a path cannot be determined until BRPC procedures are completed.
6. Network Topology Specific Considerations
In some specific network topologies the schemes for setting up
diverse LSPs could be significantly simplified.
For example, consider the L1VPN or GMPLS UNI case. This may be
viewed as a linear sequence of three domains where the first and last
domains contain only a single node each. In such a scenario, no BRPC
procedures are needed, because there is no need for path computation
in the first and last domains even if the source and destination
nodes are multi-homed.
7. Addressing Considerations
All of the schemes described in this document are applicable when a
single address space is used across all domains.
There may also be cases where private address spaces are used within
some of the domains. This problem is similar to the use of domain
confidentiality since the ERO and RRO are meaningless outside a
domain even if they are available, and the problem can be solved
using the same techniques.
8. Note on SRLG Diversity
The schemes described in this document are applicable when the nodes
and links in different domains belong to different SRLGs, which is
normally the case.
However, it is possible that nodes or links in different domains
belong to the same SRLG. That is, an SRLG may span domain
boundaries. In such cases, in order to establish SRLG diverse LSPs,
several considerations are needed:
- Record of the SRLGs used by the working LSP.
- Indication of a set of SRLGs to exclude in the computation of the
recovery LSP's path.
In this case, there is a conflict between any requirement for domain
confidentiality, and the requirement for SRLG diversity. One of the
requirements must be compromised.
Furthermore, SRLG IDs may be assigned independently in each domain,
and might not have global meaning. In such a scenario, some mapping
functions are necessary, similar to the mapping of other TE
parameters mentioned in Section 1.2.
9. Security Considerations
The core protocols used to achieve the procedures described in this
document are RSVP-TE and PCEP. These protocols include policy and
authentication capabilities as described in [RFC3209] and [PCEP].
Furthermore, these protocols may be operated using more advanced
security features such as IPsec [RFC4301] and TLS [RFC4346].
Security may be regarded as particularly important in inter-domain
deployments and serious consideration should be given to applying the
available security techniques, as described in the documents
referenced above and as set out in [RFC4726].
Additional discussion of security considerations for MPLG/GMPLS
networks can be found in [SECURITY-FW].
This document does not of itself require additional security measures
and does not modify the trust model implicit in the protocols used.
Note, however, that domain confidentiality (that is the
confidentiality of the topology and path information from within any
one domain) is an important consideration in this document, and a
significant number of the mechanisms described in this document are
designed to preserve domain confidentiality.
10.1. Normative References
[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.
[RFC4216] Zhang, R., Ed., and J.-P. Vasseur, Ed., "MPLS
Inter-Autonomous System (AS) Traffic Engineering
(TE) Requirements", RFC 4216, November 2005.
[RFC4427] Mannie, E., Ed., and D. Papadimitriou, Ed.,
"Recovery (Protection and Restoration) Terminology
for Generalized Multi-Protocol Label Switching
(GMPLS)", RFC 4427, March 2006.
[RFC4726] Farrel, A., Vasseur, J.-P., and A. Ayyangar, "A
Framework for Inter-Domain Multiprotocol Label
Switching Traffic Engineering", RFC 4726, November
10.2. Informative References
[RFC3812] Srinivasan, C., Viswanathan, A., and T. Nadeau,
"Multiprotocol Label Switching (MPLS) Traffic
Engineering (TE) Management Information Base (MIB)",
RFC 3812, June 2004.
[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.
[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.
[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
[RFC4301] Kent, S. and K. Seo, "Security Architecture for the
Internet Protocol", RFC 4301, December 2005.
[RFC4346] Dierks, T. and E. Rescorla, "The Transport Layer
Security (TLS) Protocol Version 1.1", RFC 4346,
[RFC4428] Papadimitriou, D., Ed., and E. Mannie, Ed.,
"Analysis of Generalized Multi-Protocol Label
Switching (GMPLS)-based Recovery Mechanisms
(including Protection and Restoration)", RFC 4428,
[RFC4655] Farrel, A., Vasseur, J.-P., and J. Ash, "A Path
Computation Element (PCE)-Based Architecture", RFC
4655, August 2006.
[RFC4802] Nadeau, T., Ed., and A. Farrel, Ed., "Generalized
Multiprotocol Label Switching (GMPLS) Traffic
Engineering Management Information Base", RFC 4802,
[RFC4847] Takeda, T., Ed., "Framework and Requirements for
Layer 1 Virtual Private Networks", RFC 4847, April
[RFC4872] Lang, J., Ed., Rekhter, Y., Ed., and D.
Papadimitriou, Ed., "RSVP-TE Extensions in Support
of End-to-End Generalized Multi-Protocol Label
Switching (GMPLS) Recovery", RFC 4872, May 2007.
[RFC4873] Berger, L., Bryskin, I., Papadimitriou, D., and A.
Farrel, "GMPLS Segment Recovery", RFC 4873, May
[RFC4874] Lee, CY., Farrel, A., and S. De Cnodder, "Exclude
Routes - Extension to Resource ReserVation
Protocol-Traffic Engineering (RSVP-TE)", RFC 4874,
[RFC4920] Farrel, A., Ed., Satyanarayana, A., Iwata, A.,
Fujita, N., and G. Ash, "Crankback Signaling
Extensions for MPLS and GMPLS RSVP-TE", RFC 4920,
[RFC5150] Ayyangar, A., Kompella, K., Vasseur, JP., and A.
Farrel, "Label Switched Path Stitching with
Generalized Multiprotocol Label Switching Traffic
Engineering (GMPLS TE)", RFC 5150, February 2008.
[RFC5151] Farrel, A., Ed., Ayyangar, A., and JP. Vasseur,
"Inter-Domain MPLS and GMPLS Traffic Engineering --
Resource Reservation Protocol-Traffic Engineering
(RSVP-TE) Extensions", RFC 5151, February 2008.
[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
[BRPC] Vasseur, JP., Ed., Zhang, R., Bitar, N., and JL. Le
Roux, "A Backward Recursive PCE-Based Computation
(BRPC) Procedure to Compute Shortest Inter-Domain
Traffic Engineering Label Switched Paths", Work in
Progress, April 2008.
[PCE-PATH-KEY] Bradford, R., Vasseur, JP., and A. Farrel,
"Preserving Topology Confidentiality in Inter-Domain
Path Computation Using a Key-Based Mechanism", Work
in Progress, May 2008.
[PCEP] Vasseur, JP., Ed., and JL. Le Roux, Ed., "Path
Computation Element (PCE) Communication Protocol
(PCEP)", Work in Progress, March 2008.
[PCEP-XRO] Oki, E., Takeda, T., and A. Farrel, "Extensions to
the Path Computation Element Communication Protocol
(PCEP) for Route Exclusions", Work in Progress, July
[RSVP-PATH-KEY] Bradford, R., Vasseur, JP., and A. Farrel, "RSVP
Extensions for Path Key Support", Work in Progress,
[SECURITY-FW] Fang, L., Ed., " Security Framework for MPLS and
GMPLS Networks", Work in Progress, July 2008.
The authors would like to thank Eiji Oki, Ichiro Inoue, Kazuhiro
Fujihara, Dimitri Papadimitriou, and Meral Shirazipour for valuable
comments. Deborah Brungard provided useful advice about the text.
NTT Network Service Systems Laboratories, NTT Corporation
Musashino-Shi, Tokyo 180-8585 Japan
EMail : email@example.com
NTT Communications Corporation
Tokyo Opera City Tower 3-20-2 Nishi Shinjuku, Shinjuku-ku
Tokyo 163-1421, Japan
Old Dog Consulting
Cisco Systems, Inc.
300 Beaver Brook Road
Boxborough, MA 01719
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.
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
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