Network Working Group N. Bitar, Ed.
Request for Comments: 5254 Verizon
Category: Informational M. Bocci, Ed.
L. Martini, Ed.
Cisco Systems, Inc.
October 2008Requirements for Multi-Segment Pseudowire Emulation Edge-to-Edge (PWE3)
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.
This document describes the necessary requirements to allow a service
provider to extend the reach of pseudowires across multiple domains.
These domains can be autonomous systems under one provider
administrative control, IGP areas in one autonomous system, different
autonomous systems under the administrative control of two or more
service providers, or administratively established pseudowire
This document specifies requirements for extending pseudowires across
more than one packet switched network (PSN) domain and/or more than
one PSN tunnel. These pseudowires are called multi-segment
pseudowires (MS-PWs). Requirements for single-segment pseudowires
(SS-PWs) that extend edge to edge across only one PSN domain are
specified in [RFC3916]. This document is not intended to invalidate
any part of [RFC3985].
This document specifies additional requirements that apply to MS-PWs.
These requirements do not apply to PSNs that only support SS-PWs.
The following three figures describe the reference models that are
derived from [RFC3985] to support PW emulated services.
|<-------------- Emulated Service ---------------->|
| |<------- Pseudowire ------->| |
| | | |
| | |<-- PSN Tunnel -->| | |
| PW End V V V V PW End |
V Service +----+ +----+ Service V
+-----+ | | PE1|==================| PE2| | +-----+
| |----------|............PW1.............|----------| |
| CE1 | | | | | | | | CE2 |
| |----------|............PW2.............|----------| |
+-----+ ^ | | |==================| | | ^ +-----+
^ | +----+ +----+ | | ^
| | Provider Edge 1 Provider Edge 2 | |
| | | |
Customer | | Customer
Edge 1 | | Edge 2
Attachment Circuit (AC) Attachment Circuit (AC)
Native service Native service
Figure 1: PWE3 Reference Configuration
Figure 1 shows the PWE3 reference architecture [RFC3985]. This
architecture applies to the case where a PSN tunnel extends between
two edges of a single PSN domain to transport a PW with endpoints at
Native |<--------Multi-Segment Pseudowire----->| Native
Service | PSN PSN | Service
(AC) | |<-Tunnel->| |<-Tunnel->| | (AC)
| V V 1 V V 2 V V |
| +-----+ +-----+ +---- + |
+---+ | |T-PE1|==========|S-PE1|==========|T-PE2| | +---+
| |---------|........PW1.......... |...PW3..........|---|----| |
|CE1| | | | | | | | | |CE2|
| |---------|........PW2...........|...PW4..........|--------| |
+---+ | | |==========| |==========| | | +---+
^ +-----+ +-----+ +-----+ ^
| Provider Edge 1 ^ Provider Edge 3 |
| | |
| | |
| PW switching point |
|<------------------- Emulated Service ------------------->|
Figure 2: PW Switching Reference ModelFigure 2 extends this architecture to show a multi-segment case.
Terminating PE1 (T-PE1) and Terminating PE3 (T-PE3) provide PWE3
service to CE1 and CE2. These PEs terminate different PSN tunnels,
PSN Tunnel 1 and PSN Tunnel 2, and may reside in different PSN or
pseudowire domains. One PSN tunnel extends from T-PE1 to S-PE1
across PSN1, and a second PSN tunnel extends from S-PE1 to T-PE2
PWs are used to connect the Attachment circuits (ACs) attached to
T-PE1 to the corresponding ACs attached to T-PE2. Each PW on PSN
tunnel 1 is switched to a PW in the tunnel across PSN2 at S-PE1 to
complete the multi-segment PW (MS-PW) between T-PE1 and T-PE2. S-PE1
is therefore the PW switching point and will be referred to as the PW
switching provider edge (S-PE). PW1 and PW3 are segments of the same
MS-PW while PW2 and PW4 are segments of another pseudowire. PW
segments of the same MS-PW (e.g., PW1 and PW3) MAY be of the same PW
type or different types, and PSN tunnels (e.g., PSN Tunnel 1 and PSN
Tunnel 2) can be the same or different technology. This document
requires support for MS-PWs with segments of the same PW type only.
An S-PE switches an MS-PW from one segment to another based on the PW
identifiers (e.g., PW label in case of MPLS PWs). In Figure 2, the
domains that PSN Tunnel 1 and PSN Tunnel 2 traverse could be IGP
areas in the same IGP network or simply PWE3 domains in a single flat
IGP network, for instance.
| AS AS |
AC | |<----1---->| |<----2--->| | AC
| V V V V V V |
| +----+ +-----+ +----+ +----+ |
+----+ | | |=====| |=====| |=====| | | +----+
| |-------|.....PW1..........PW2.........PW3.....|-------| |
| CE1| | | | | | | | | | | |CE2 |
+----+ | | |=====| |=====| |=====| | | +----+
^ +----+ +-----+ +----+ +----+ ^
| T-PE1 S-PE2 S-PE3 T-PE4 |
| ^ ^ |
| | | |
| PW switching points |
|<------------------- Emulated Service --------------->|
Figure 3: PW Switching Inter-Provider Reference Model
Note that although Figure 2 only shows a single S-PE, a PW may
transit more than one S-PEs along its path. For instance, in the
multi-AS case shown in Figure 3, there can be an S-PE (S-PE2) at the
border of one AS (AS1) and another S-PE (S-PE3) at the border of the
other AS (AS2). An MS-PW that extends from the edge of one AS (T-
PE1) to the edge of the other AS (T-PE4) is composed of three
segments: (1) PW1, a segment in AS1, (2) PW2, a segment between the
two border routers (S-PE2 and S-PE3) that are switching PEs, and (3)
PWE3, a segment in AS2. AS1 and AS2 could belong to the same
provider (e.g., AS1 could be an access network or metro transport
network, and AS2 could be an MPLS core network) or to two different
providers (e.g., AS1 for Provider 1 and AS2 for Provider 2).
RFC 3985 [RFC3985] provides terminology for PWE3. The following
additional terminology is defined for multi-segment pseudowires:
- PW Terminating Provider Edge (T-PE). A PE where the
customer-facing attachment circuits (ACs) are bound to a PW
forwarder. A Terminating PE is present in the first and last
segments of an MS-PW. This incorporates the functionality of a
PE as defined in RFC 3985.
- Single-Segment Pseudowire (SS-PW). A PW setup directly between
two PE devices. Each direction of an SS-PW traverses one PSN
tunnel that connects the two PEs.
- Multi-Segment Pseudowire (MS-PW). A static or dynamically
configured set of two or more contiguous PW segments that
behave and function as a single point-to-point PW. Each end of
an MS-PW by definition MUST terminate on a T-PE.
- PW Segment. A single-segment or a part of a multi-segment PW,
which is set up between two PE devices, T-PEs and/or S-PEs.
- PW Switching Provider Edge (S-PE). A PE capable of switching
the control and data planes of the preceding and succeeding PW
segments in an MS-PW. The S-PE terminates the PSN tunnels
transporting the preceding and succeeding segments of the MS-
PW. It is therefore a PW switching point for an MS-PW. A PW
switching point is never the S-PE and the T-PE for the same
MS-PW. A PW switching point runs necessary protocols to set up
and manage PW segments with other PW switching points and
2.1. Specification of Requirements
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119].
3. Use Cases
PWE3 defines the signaling and encapsulation techniques for
establishing SS-PWs between a pair of terminating PEs (T-PEs), and in
the vast majority of cases, this will be sufficient. MS-PWs may be
useful in the following situations:
-i. Inter-Provider PWs: An Inter-Provider PW is a PW that extends
from a T-PE in one provider domain to a T-PE in another
-ii. It may not be possible, desirable, or feasible to establish a
direct PW control channel between the T-PEs, residing in
different provider networks, to set up and maintain PWs. At a
minimum, a direct PW control channel establishment (e.g.,
targeted LDP session) requires knowledge of and reachability
to the remote T-PE IP address. The local T-PE may not have
access to this information due to operational or security
constraints. Moreover, an SS-PW would require the existence
of a PSN tunnel between the local T-PE and the remote T-PE.
It may not be feasible or desirable to extend single,
contiguous PSN tunnels between T-PEs in one domain and T-PEs
in another domain for security and/or scalability reasons or
because the two domains may be using different PSN
-iii. MS-PW setup, maintenance, and forwarding procedures must
satisfy requirements placed by the constraints of a
multi-provider environment. An example is the inter-AS L2VPN
scenario where the T-PEs reside in different provider networks
(ASs) and it is the current practice to MD5-key all control
traffic exchanged between two networks. An MS-PW allows the
providers to confine MD5 key administration for the LDP
session to just the PW switching points connecting the two
-iv. PSN Interworking: PWE3 signaling protocols and PSN types may
differ in different provider networks. The terminating PEs
may be connected to networks employing different PW signaling
and/or PSN protocols. In this case, it is not possible to use
an SS-PW. An MS-PW with the appropriate interworking
performed at the PW switching points can enable PW
connectivity between the terminating PEs in this scenario.
-v. Traffic Engineered PSN Tunnels and bandwidth-managed PWs:
There is a requirement to deploy PWs edge to edge in large
service provider networks. Such networks typically encompass
hundreds or thousands of aggregation devices at the edge, each
of which would be a PE. Furthermore, there is a requirement
that these PWs have explicit bandwidth guarantees. To satisfy
these requirements, the PWs will be tunneled over PSN
TE-tunnels with bandwidth constraints. A single-segment
pseudowire architecture would require that a full mesh of PSN
TE-tunnels be provisioned to allow PWs to be established
between all PEs. Inter-provider PWs riding traffic engineered
tunnels further add to the number of tunnels that would have
to be supported by the PEs and the core network as the total
number of PEs increases.
In this environment, there is a requirement either to support
a sparse mesh of PSN TE-tunnels and PW signaling adjacencies,
or to partition the network into a number of smaller PWE3
domains. In either case, a PW would have to pass through more
than one PSN tunnel hop along its path. An objective is to
reduce the number of tunnels that must be supported, and thus
the complexity and scalability problem that may arise.
-vi. Pseudowires in access/metro networks: Service providers wish
to extend PW technology to access and metro networks in order
to reduce maintenance complexity and operational costs.
Today's access and metro networks are either legacy (Time
Division Multiplexed (TDM), Synchronous Optical
Network/Synchronous Digital Hierarchy (SONET/SDH), or Frame
Relay/Asynchronous Transfer Mode (ATM)), Ethernet, or IP
Due to these architectures, circuits (e.g., Ethernet Virtual
Circuits (EVCs), ATM VCs, TDM circuits) in the access/metro
are traditionally handled as attachment circuits, in their
native format, to the edge of the IP-MPLS network where the PW
starts. This combination requires multiple separate access
networks and complicates end-to-end control, provisioning, and
maintenance. In addition, when a TDM or SONET/SDH access
network is replaced with a packet-based infrastructure,
expenses may be lowered due to moving statistical multiplexing
closer to the end-user and converging multiple services onto a
single access network.
Access networks have a number of properties that impact the
application of PWs. For example, there exist access
mechanisms where the PSN is not of an IETF specified type, but
uses mechanisms compatible with those of PWE3 at the PW layer.
Here, use case (iv) may apply. In addition, many networks
consist of hundreds or thousands of access devices. There is
therefore a desire to support a sparse mesh of PW signaling
adjacencies and PSN tunnels. Use case (v) may therefore
apply. Finally, access networks also tend to differ from core
networks in that the access PW setup and maintenance mechanism
may only be a subset of that used in the core.
Using the MS-PWs, access and metro network elements need only
maintain PW signaling adjacencies with the PEs to which they
directly connect. They do not need PW signaling adjacencies
with every other access and metro network device. PEs in the
PSN backbone, in turn, maintain PW signaling adjacencies among
each other. In addition, a PSN tunnel is set up between an
access element and the PE to which it connects. Another PSN
tunnel needs to be established between every PE pair in the
PSN backbone. An MS-PW may be set up from one access network
element to another access element with three segments: (1)
access-element - PSN-PE, (2) PSN-PE to PSN-PE, and (3) PSN-PE
to access element. In this MS-PW setup, access elements are
T-PEs while PSN-PEs are S-PEs. It should be noted that the
PSN backbone can be also segmented into PWE3 domains resulting
in more segments per PW.
3.1. Multi-Segment Pseudowire Setup Mechanisms
This requirements document assumes that the above use cases are
realized using one or more of the following mechanisms:
-i. Static Configuration: The switching points (S-PEs), in
addition to the T-PEs, are manually provisioned for each
-ii. Pre-Determined Route: The PW is established along an
administratively determined route using an end-to-end
signaling protocol with automated stitching at the S-PEs.
-iii. Signaled Dynamic Route: The PW is established along a
dynamically determined route using an end-to-end signaling
protocol with automated stitching at the S-PEs. The route is
selected with the aid of one or more dynamic routing
Note that we define the PW route to be the set of S-PEs through which
an MS-PW will pass between a given pair of T-PEs. PSN tunnels along
that route can be explicitly specified or locally selected at the
S-PEs and T-PEs. The routing of the PSN tunnels themselves is
outside the scope of the requirements specified in this document.
4. Multi-Segment Pseudowire Requirements
The following sections detail the requirements that the above use
cases put on the MS-PW setup mechanisms.
4.1. All Mechanisms
The following generic requirements apply to the three MS-PW setup
mechanisms defined in the previous section.
-i. If MS-PWs are tunneled across a PSN that only supports SS-PWs,
then only the requirements of [RFC3916] apply to that PSN.
The fact that the overlay is carrying MS-PWs MUST be
transparent to the routers in the PSN.
-ii. The PWs MUST remain transparent to the P-routers. A P-router
is not an S-PE or an T-PE from the MS-PW architecture
viewpoint. P-routers provide transparent PSN transport for
PWs and MUST not have any knowledge of the PWs traversing
-iii. The MS-PWs MUST use the same encapsulation modes specified for
-iv. The MS-PWs MUST be composed of SS-PWs.
-v. An MS-PW MUST be able to pass across PSNs of all technologies
supported by PWE3 for SS-PWs. When crossing from one PSN
technology to another, an S-PE must provide the necessary PSN
interworking functions in that case.
-vi. Both directions of a PW segment MUST terminate on the same
-vii. S-PEs MAY only support switching PWs of the same PW type. In
this case, the PW type is transparent to the S-PE in the
forwarding plane, except for functions needed to provide for
interworking between different PSN technologies.
-viii. Solutions MAY provide a way to prioritize the setup and
maintenance process among PWs.
Mechanisms to protect an MS-PW when an element on the existing path
of an MS-PW fails MUST be provided. These mechanisms will depend on
the MS-PW setup. The following are the generic resiliency
requirements that apply to all MS-PW setup mechanisms:
-i. Configuration and establishment of a backup PW to a primary PW
SHOULD be supported. Mechanisms to perform a switchover from
a primary PW to a backup PW upon failure detection SHOULD be
-ii. The ability to configure an end-to-end backup PW path for a
primary PW path SHOULD be supported. The primary and backup
paths may be statically configured, statically specified for
signaling, or dynamically selected via dynamic routing
depending on the MS-PW establishment mechanism. Backup and
primary paths should have the ability to traverse separate
S-PEs. The backup path MAY be signaled at configuration time
or after failure.
-iii. The ability to configure a primary PW and a backup PW with a
different T-PE from the primary SHOULD be supported.
-iv. Automatic Mechanisms to perform a fast switchover from a
primary PW to a backup PW upon failure detection SHOULD be
-v. A mechanism to automatically revert to a primary PW from a
backup PW MAY be provided. When provided, it MUST be
4.1.3. Quality of Service
Pseudowires are intended to support emulated services (e.g., TDM and
ATM) that may have strict per-connection quality-of-service (QoS)
requirements. This may include either absolute or relative
guarantees on packet loss, delay, and jitter. These guarantees are,
in part, delivered by reserving sufficient network resources (e.g.,
bandwidth), and by providing appropriate per-packet treatment (e.g.,
scheduling priority and drop precedence) throughout the network.
For SS-PWs, a traffic engineered PSN tunnel (i.e., MPLS-TE) may be
used to ensure that sufficient resources are reserved in the
P-routers to provide QoS to PWs on the tunnel. In this case, T-PEs
MUST have the ability to automatically request the PSN tunnel
resources in the direction of traffic (e.g., admission control of PWs
onto the PSN tunnel and accounting for reserved bandwidth and
available bandwidth on the tunnel). In cases where the tunnel
supports multiple classes of service (CoS) (e.g., E-LSP), bandwidth
management is required per CoS.
For MS-PWs, each S-PE maps a PW segment to a PSN tunnel. Solutions
MUST enable S-PEs and T-PEs to automatically bind a PW segment to a
PSN tunnel based on CoS and bandwidth requirements when these
attributes are specified for a PW. Solutions SHOULD also provide the
capability of binding a PW segment to a tunnel as a matter of policy
configuration. S-PEs and T-PEs must be capable of automatically
requesting PSN tunnel resources per CoS.
S-PEs and T-PEs MUST be able to associate a CoS marking (e.g., EXP
field value for MPLS PWs) with PW PDUs. CoS marking in the PW PDUs
affects packet treatment. The CoS marking depends on the PSN
technology. Thus, solutions must enable the configuration of
necessary mapping for CoS marking when the MS-PW crosses from one PSN
technology to another. Similarly, different administrative domains
may use different CoS values to imply the same CoS treatment.
Solutions MUST provide the ability to define CoS marking maps on
S-PEs at administrative domain boundaries to translate from one CoS
value to another as a PW PDU crosses from one domain to the next.
[RFC3985] requires PWs to respond to path congestion by reducing
their transmission rate. Alternatively, RFC 3985 permits PWs that do
not have a congestion control mechanism to transmit using explicitly
reserved capacity along a provisioned path. Because MS-PWs are a
type of PW, this requirement extends to them as well. RFC 3985
applied to MS-PWs consequently requires that MS-PWs employ a
congestion control mechanism that is effective across an MS path, or
requires an explicit provisioning action that reserves sufficient
capacity in all domains along the MS path before the MS-PW begins
transmission. S-PEs are therefore REQUIRED to reject attempts to
establish MS-PW segments for PW types that either do not utilize an
appropriate congestion control scheme or when resources that are
sufficient to support the transmission rate of the PW cannot be
reserved along the path.
4.1.4. Congestion Control
[RFC3985] requires all PWs to respond to congestion, in order to
conform to [RFC2914]. In the absence of a well-defined congestion
control mechanism, [RFC3985] permits PWs to be carried across paths
that have been provisioned such that the traffic caused by PWs has no
harmful effect on concurrent traffic that shares the path, even under
congestion. These requirements extend to the MS-PWs defined in this
Path provisioning is frequently performed through QoS reservation
protocols or network management protocols. In the case of SS-PWs,
which remain within a single administrative domain, a number of
existing protocols can provide this provisioning functionality. MS-
PWs, however, may transmit across network domains that are under the
control of multiple entities. QoS provisioning across such paths is
inherently more difficult, due to the required inter-domain
interactions. It is important to note that these difficulties do not
invalidate the requirement to provision path capacity for MS-PW use.
Each domain MUST individually implement a method to control
congestion. This can be by QoS reservation, or other congestion
control method. MS-PWs MUST NOT transmit across unprovisioned, best
effort, paths in the absence of other congestion control schemes, as
required by [RFC3985].
Solutions MUST enable S-PEs and T-PEs on the path of an MS-PW to
notify other S-PEs and T-PEs on that path of congestion, when it
occurs. Congestion may be indicated by queue length, packet loss
rate, or bandwidth measurement (among others) crossing a respective
threshold. The action taken by a T-PE that receives a notification
of congestion along the path of one of its PWs could be to re-route
the MS-PW to an alternative path, including an alternative T-PE if
available. If a PE, or an S-PE has knowledge that a particular link
or tunnel is experiencing congestion, it MUST not set up any new
MS-PW that utilize that link or tunnel. Some PW types, such as TDM
PWs, are more sensitive to congestion than others. The reaction to a
congestion notification MAY vary per PW type.
4.1.5. Generic Requirements for MS-PW Setup Mechanisms
The MS-PW setup mechanisms MUST accommodate the service provider's
practices, especially in relation to security, confidentiality of SP
information, and traffic engineering. Security and confidentiality
are especially important when the MS-PWs are set up across autonomous
systems in different administrative domains. The following are
generic requirements that apply to the three MS-PW setup mechanisms
-i. The ability to statically select S-PEs and PSN tunnels on a PW
path MUST be provided. Static selection of S-PEs is by
definition a requirement for the static configuration and
signaled/static route setup mechanisms. This requirement
satisfies the need for forcing an MS-PW to traverse specific
S-PEs to enforce service provider security and administrative
-ii. Solutions SHOULD minimize the amount of configuration needed
to set up an MS-PW.
-iii. Solutions should support different PW setup mechanisms on the
same T-PE, S-PE, and PSN network.
-iv. Solutions MUST allow T-PEs to simultaneously support use of
SS-PW signaling mechanisms as specified in [RFC4447], as well
as MS-PW signaling mechanisms.
-v. Solutions MUST ensure that an MS-PW will be set up when a path
that satisfies the PW constraints for bandwidth, CoS, and
other possible attributes does exist in the network.
-vi. Solutions must clearly define the setup procedures for each
mechanism so that an MS-PW setup on T-PEs can be interpreted
as successful only when all PW segments are successfully set
-vii. Admission control to the PSN tunnel needs to be performed
against available resources, when applicable. This process
MUST be performed at each PW segment comprising the MS-PW. PW
admission control into a PSN tunnel MUST be configurable.
-viii. In case the PSN tunnel lacks the resources necessary to
accommodate the new PW, an attempt to signal a new PSN tunnel,
or increase the capacity of the existing PSN tunnel MAY be
made. If the expanded PSN tunnel fails to set up, the PW MUST
fail to set up.
-ix. The setup mechanisms must allow the setup of a PW segment
between two directly connected S-PEs without the existence of
a PSN tunnel. This requirement allows a PW segment to be set
up between two (Autonomous System Border Routers (ASBRs) when
the MS-PW crosses AS boundaries without the need for
configuring and setting up a PSN tunnel. In this case,
admission control must be done, when enabled, on the link
between the S-PEs.
An objective of MS-PWs is to provide support for the following
-i. MS-PWs MUST be able to traverse multiple service provider
-ii. MS-PWs MUST be able to traverse multiple autonomous systems
within the same administrative domain.
-iii. MS-PWs MUST be able to traverse multiple autonomous systems
belonging to different administrative domains.
-iv. MS-PWs MUST be able to support any hybrid combination of the
aforementioned connectivity scenarios, including both PW
transit and termination in a domain.
4.2. Statically Configured MS-PWs
When the MS-PW segments are statically configured, the following
requirements apply in addition to the generic requirements previously
There are no additional requirements on the architecture.
Solutions should allow for the static configuration of MPLS labels
for MPLS-PW segments and the cross-connection of these labels to
preceding and succeeding segments. This is especially useful when an
MS-PW crosses provider boundaries and two providers do not want to
run any PW signaling protocol between them. A T-PE or S-PE that
allows the configuration of static labels for MS-PW segments should
also simultaneously allow for dynamic label assignments for other
MS-PW segments. It should be noted that when two interconnected
S-PEs do not have signaling peering for the purpose of setting up
MS-PW segments, they should have in-band PW Operations and
Maintenance (OAM) capabilities that relay PW or attachment circuit
defect notifications between the adjacent S-PEs.
The solution should allow for the protection of a PW segment, a
contiguous set of PW segments, as well as the end-to-end path. The
primary and protection segments must share the same segment
endpoints. Solutions should allow for having the backup paths set up
prior to the failure or as a result of failure. The choice should be
made by configuration. When resources are limited and cannot satisfy
all PWs, the PWs with the higher setup priorities should be given
preference when compared with the setup priorities of other PWs being
set up or the holding priorities of existing PWs.
Solutions should strive to minimize traffic loss between T-PEs.
4.2.4. Quality of Service
The CoS and bandwidth of the MS-PW must be configurable at T-PEs and
4.3. Signaled PW Segments
When the MS-PW segments are dynamically signaled, the following
requirements apply in addition to the generic requirements previously
defined. The signaled MS-PW segments can be on the path of a
statically configured MS-PW, signaled/statically routed MS-PW, or
signaled/dynamically routed MS-PW.
There are four different mechanisms that are defined to setup SS-PWs:
-i. Static set up of the SS-PW (MPLS or L2TPv3 forwarding)
-ii. LDP using PWid Forwarding Equivalence Class (FEC) 128
-iii. LDP using the generalized PW FEC 129
The MS-PW setup mechanism MUST be able to support PW segments
signaled with any of the above protocols; however, the specification
of which combinations of SS-PW signaling protocols are supported by a
specific implementation is outside the scope of this document.
For the signaled/statically routed and signaled/dynamically routed
MS-PW setup mechanisms, the following requirements apply in addition
to the generic requirements previously defined.
There are no additional requirements on the architecture.
Solutions should allow for the signaling of a protection path for a
PW segment, sequence of segments, or end-to-end path. The protection
and primary paths for the protected segment(s) share the same
respective segments endpoints. When admission control is enabled,
systems must be careful not to double account for bandwidth
allocation at merged points (e.g., tunnels). Solutions should allow
for having the backup paths set up prior to the failure or as a
result of failure. The choice should be made by configuration at the
endpoints of the protected path. When resources are limited and
cannot satisfy all PWs, the PWs with the higher setup priorities
should be given preference when compared with the setup priorities of
other PWs being set up or the holding priorities of existing PWs.
Procedures must allow for the primary and backup paths to be diverse.
4.3.3. Quality of Service
When the T-PE attempts to signal an MS-PW, the following capability
-i. Signaling must be able to identify the CoS associated with an
-ii. Signaling must be able to carry the traffic parameters for an
MS-PW per CoS. Traffic parameters should be based on existing
INTSERV definitions and must be used for admission control
when admission control is enabled.
-iii. The PW signaling MUST enable separate traffic parameter values
to be specified for the forward and reverse directions of the
-iv. PW traffic parameter representations MUST be the same for all
types of MS-PWs.
-v. The signaling protocol must be able to accommodate a method to
prioritize the PW setup and maintenance operation among PWs.
See the requirements for "Resiliency" above.
4.3.5. Additional Requirements on Signaled MS-PW Setup Mechanisms
The following are further requirements on signaled MS-PW setup
-i. The signaling procedures MUST be defined such that the setup
of an MS-PW is considered successful if all segments of the
MS-PW are successfully set up.
-ii. The MS-PW path MUST have the ability to be dynamically set up
between the T-PEs by provisioning only the T-PEs.
-iii. Dynamic MS-PW setup requires that a unique identifier be
associated with a PW and be carried in the signaling message.
That identifier must contain sufficient information to
determine the path to the remote T-PE through intermediate
-iv. In a single-provider domain, it is natural to have the T-PE
identified by one of its IP addresses. This may also apply
when an MS-PW is set up across multiple domains operated by
the same provider. However, some service providers have
security and confidentiality policies that prevent them from
advertising reachability to routers in their networks to other
providers (reachability to an ASBR is an exception). Thus,
procedures MUST be provided to allow dynamic set up of MS-PWs
under these conditions.
4.4. Signaled PW / Dynamic Route
The following requirements apply, in addition to those in Sections
4.1 and 4.3, when both dynamic signaling and dynamic routing are
There are no additional architectural requirements.
The PW routing function MUST support dynamic re-routing around
failure points when segments are set up using the dynamic setup
4.4.3. Quality of Service
There are no additional QoS requirements.
The following are requirements associated with dynamic route
selection for an MS-PW:
-i. Routing must enable S-PEs and T-PEs to discover S-PEs on the
path to a destination T-PE.
-ii. The MS-PW routing function MUST have the ability to
automatically select the S-PEs along the MS-PW path. Some of
the S-PEs MAY be statically selected and carried in the
signaling to constrain the route selection process.
-iii. The PW routing function MUST support re-routing around
failures that occur between the statically configured segment
endpoints. This may be done by choosing another PSN tunnel
between the two segment endpoints or setting up an alternative
-iv. Routing protocols must be able to advertise reachability
information of attachment circuit (AC) endpoints. This
reachability information must be consistent with the AC
identifiers carried in signaling.
5. Operations and Maintenance (OAM)
OAM mechanisms for the attachment circuits are defined in the
specifications for PW emulated specific technologies (e.g., ITU-T
I.610 [i610] for ATM). These mechanisms enable, among other things,
defects in the network to be detected, localized, and diagnosed.
They also enable communication of PW defect states on the PW
attachment circuit. Note that this document uses the term OAM as
Operations and Management as per ITU-T I.610.
The interworking of OAM mechanisms for SS-PWs between ACs and PWs is
defined in [PWE3-OAM]. These enable defect states to be propagated
across a PWE3 network following the failure and recovery from faults.
OAM mechanisms for MS-PWs MUST provide at least the same capabilities
as those for SS-PWs. In addition, it should be possible to support
both segment and end-to-end OAM mechanisms for both defect
notifications and connectivity verification in order to allow defects
to be localized in a multi-segment network. That is, PW OAM segments
can be T-PE to T-PE, T-PE to S-PE, or S-PE to S-PE.
The following requirements apply to OAM for MS-PWs:
-i. Mechanisms for PW segment failure detection and notification
to other segments of an MS-PW MUST be provided.
-ii. MS-PW OAM SHOULD be supported end-to-end across the network.
-iii. Single ended monitoring SHOULD be supported for both
directions of the MS-PW.
-iv. SS-PW OAM mechanisms (e.g., [RFC5085]) SHOULD be extended to
support MS-PWs on both an end-to-end basis and segment basis.
-v. All PE routers along the MS-PW MUST agree on a common PW OAM
mechanism to use for the MS-PW.
-vi. At the S-PE, defects on an PSN tunnel MUST be propagated to
all PWs that utilize that particular PSN tunnel.
-vii. The directionality of defect notifications MUST be maintained
across the S-PE.
-viii. The S-PE SHOULD be able to behave as a segment endpoint for PW
-ix. The S-PE MUST be able to pass T-PE to T-PE PW OAM messages
-x. Performance OAM is required for both MS-PWs and SS-PWs to
measure round-trip delay, one-way delay, jitter, and packet
6. Management of Multi-Segment Pseudowires
Each PWE3 approach that uses MS-PWs SHOULD provide some mechanisms
for network operators to manage the emulated service. Management
mechanisms for MS-PWs MUST provide at least the same capabilities as
those for SS-PWs, as defined in [RFC3916].
It SHOULD also be possible to manage the additional attributes for
MS-PWs. Since the operator that initiates the establishment of an
MS-PW may reside in a different PSN domain from the S-PEs and one of
the T-PEs along the path of the MS-PW, mechanisms for the remote
management of the MS-PW SHOULD be provided.
The following additional requirements apply:
6.1. MIB Requirements
-i. MIB Tables MUST be designed to facilitate configuration and
provisioning of the MS-PW at the S-PEs and T-PEs.
-ii. The MIB(s) MUST facilitate inter-PSN configuration and
monitoring of the ACs.
6.2. Management Interface Requirements
-i. Mechanisms MUST be provided to enable remote management of an
MS-PW at an S-PE or T-PE. It SHOULD be possible for these
mechanisms to operate across PSN domains. An example of a
commonly available mechanism is the command line interface
(CLI) over a telnet session.
-ii. For security or other reasons, it SHOULD be possible to
disable the remote management of an MS-PW.
7. Security Considerations
This document specifies the requirements both for MS-PWs that can be
set up across domain boundaries administered by one or more service
providers (inter-provider MS-PWs), and for MS-PWs that are only set
up across one provider (intra-provider MS-PWs).
7.1. Inter-Provider MS-PWs
The security requirements for MS-PW setup across domains administered
by one service provider are the same as those described under
security considerations in [RFC4447] and [RFC3916]. These
requirements also apply to inter-provider MS-PWs.
In addition, [RFC4111] identifies user and provider requirements for
L2 VPNs that apply to MS-PWs described in this document. In this
section, the focus is on the additional security requirements for
inter-provider operation of MS-PWs in both the control plane and data
plane, and some of these requirements may overlap with those in
7.1.1. Data-Plane Security Requirements
By security in the "data plane", we mean protection against the
-i. Packets from within an MS-PW traveling to a PE or an AC to
which the PW is not intended to be connected, other than in a
manner consistent with the policies of the MS-PW.
-ii. Packets from outside an MS-PW entering the MS-PW, other than
in a manner consistent with the policies of the MS-PW.
MS-PWs that cross service provider (SP) domain boundaries may connect
one T-PE in a SP domain to a T-PE in another provider domain. They
may also transit other provider domains even if the two T-PEs are
under the control of one SP. Under these scenarios, there is a
chance that one or more PDUs could be falsely inserted into an MS-PW
at any of the originating, terminating, or transit domains. Such
false injection can be the result of a malicious attack or fault in
the S-PE. Solutions MAY provide mechanisms for ensuring the
end-to-end authenticity of MS-PW PDUs.
The data plane security requirements at a service provider border for
MS-PWs are similar to those for inter-provider BGP/MPLS IP Virtual
Private Networks [RFC4364]. In particular, an S-PE or T-PE SHOULD
discard a packet received from a particular neighbor over the service
provider border unless one of the following two conditions holds:
-i. Any MPLS label processed at the receiving S-PE or T-PE, such
the PSN tunnel label or the PW label has a label value that
the receiving system has distributed to that neighbor; or
-ii. Any MPLS label processed at the receiving S-PE or T-PE, such
as the PSN tunnel label or the PW label has a label value that
the receiving S-PE or T-PE has previously distributed to the
peer S-PE or T-PE beyond that neighbor (i.e., when it is known
that the path from the system to which the label was
distributed to the receiving system is via that neighbor).
One of the domains crossed by an MS-PW may decide to selectively
mirror the PDUs of an MS-PW for eavesdropping purposes. It may also
decide to selectively hijack the PDUs of an MS-PW by directing the
PDUs away from their destination. In either case, the privacy of an
MS-PW can be violated.
Some types of PWs make assumptions about the security of the
underlying PSN. The minimal security provided by an MPLS PSN might
not be sufficient to meet the security requirements expected by the
applications using the MS-PW. This document does not place any
requirements on protecting the privacy of an MS-PW PDU via
encryption. However, encryption may be required at a higher layer in
the protocol stack, based on the application or network requirements.
The data plane of an S-PE at a domain boundary MUST be able to police
incoming MS-PW traffic to the MS-PW traffic parameters or to an
administratively configured profile. The option to enable/disable
policing MUST be provided to the network administrator. This is to
ensure that an MS-PW or a group of MS-PWs do not grab more resources
than they are allocated. In addition, the data plane of an S-PE MUST
be able to police OAM messages to a pre-configured traffic profile or
to filter out these messages upon administrative configuration.
An ingress S-PE MUST ensure that an MS-PW receives the CoS treatment
configured or signaled for that MS-PW at the S-PE. Specifically, an
S-PE MUST guard against packets marked in the exp bits or IP-header
Differentiated Services (DS) field (depending on the PSN) for a
better CoS than they should receive.
An ingress S-PE MUST be able to define per-interface or
interface-group (a group may correspond to interfaces to a peer-
provider) label space for MPLS-PWs. An S-PE MUST be configurable not
to accept labeled packets from another provider unless the bottom
label is a PW-label assigned by the S-PE on the interface on which
the packet arrived.
Data plane security considerations for SS-PWs specified in [RFC3985]
also apply to MS-PWs.
7.1.2. Control-Plane Security Requirements
An MS-PW connects two attachment circuits. It is important to make
sure that PW connections are not arbitrarily accepted from anywhere,
or else a local attachment circuit might get connected to an
arbitrary remote attachment circuit. The fault in the connection can
start at a remote T-PE or an S-PE.
Where a PW segment crosses a border between one provider and another
provider, the PW segment endpoints (S-PEs) SHOULD be on ASBRs
interconnecting the two providers. Directly interconnecting the
S-PEs using a physically secure link, and enabling signaling and
routing authentication between the S-PEs, eliminates the possibility
of receiving an MS-PW signaling message or packet from an untrusted
peer. Other configurations are possible. For example, P routers for
the PSN tunnel between the adjacent S-PEs/T-PEs may reside on the
ASBRs. In which case, the S-PEs/T-PEs MUST satisfy themselves of the
security and privacy of the path.
The configuration and maintenance protocol MUST provide a strong
authentication and control protocol data protection mechanism. This
option MUST be implemented, but it should be deployed according to
the specific PSN environment requirements. Furthermore,
authentication using a signature for each individual MS-PW setup
message MUST be available, in addition to an overall control protocol
session authentication and message validation.
Since S-PEs in different provider networks SHOULD reside at each end
of a physically secure link, or be interconnected by a limited number
of trusted PSN tunnels, each S-PE will have a trust relationship with
only a limited number of S-PEs in other ASs. Thus, it is expected
that current security mechanisms based on manual key management will
be sufficient. If deployment situations arise that require large
scale connection to S-PEs in other ASs, then a mechanism based on RFC
4107 [RFC4107] MUST be developed.
Peer authentication protects against IP address spoofing but does not
prevent one peer (S-PE or T-PE) from connecting to the wrong
attachment circuit. Under a single administrative authority, this
may be the result of a misconfiguration. When the MS-PW crosses
multiple provider domains, this may be the result of a malicious act
by a service provider or a security hole in that provider network.
Static manual configuration of MS-PWs at S-PEs and T-PEs provides a
greater degree of security. If an identification of both ends of an
MS-PW is configured and carried in the signaling message, an S-PE can
verify the signaling message against the configuration. To support
dynamic signaling of MS-PWs, whereby only endpoints are provisioned
and S-PEs are dynamically discovered, mechanisms SHOULD be provided
to configure such information on a server and to use that information
during a connection attempt for validation.
An incoming MS-PW request/reply MUST NOT be accepted unless its IP
source address is known to be the source of an "eligible" peer. An
eligible peer is an S-PE or a T-PE with which the originating S-PE or
T-PE has a trust relationship. The number of such trusted T-PEs or
S-PEs is bounded and not anticipated to create a scaling issue for
the control plane authentication mechanisms.
If a peering adjacency has to be established prior to exchanging
setup requests/responses, peering MUST only be done with eligible
peers. The set of eligible peers could be pre-configured (either as
a list of IP addresses, or as a list of address/mask combinations) or
automatically generated from the local PW configuration information.
Furthermore, the restriction of peering sessions to specific
interfaces MUST also be provided. The S-PE and T-PE MUST drop the
unaccepted signaling messages in the data path to avoid a
Denial-of-Service (DoS) attack on the control plane.
Even if a connection request appears to come from an eligible peer,
its source address may have been spoofed. Thus, means of preventing
source address spoofing must be in place. For example, if eligible
peers are in the same network, source address filtering at the border
routers of that network could eliminate the possibility of source
S-PEs that connect one provider domain to another provider domain
MUST have the capability to rate-limit signaling traffic in order to
prevent DoS attacks on the control plane. Furthermore, detection and
disposition of malformed packets and defense against various forms of
attacks that can be protocol-specific MUST be provided.
7.2. Intra-Provider MS-PWs
Security requirements for pseudowires are provided in [RFC3916].
These requirements also apply to MS-PWs.
MS-PWs are intended to enable many more PEs to provide PWE3 services
in a given service provider network than traditional SS-PWs,
particularly in access and metro environments where the PE may be
situated closer to the ultimate endpoint of the service. In order to
limit the impact of a compromise of one T-PE in a service provider
network, the data path security requirements for inter-provider
MS-PWs also apply to intra-provider MS-PWs in such cases.
The editors gratefully acknowledge the following contributors:
Dimitri Papadimitriou (Alcatel-Lucent), Peter Busschbach
(Alcatel-Lucent), Sasha Vainshtein (Axerra), Richard Spencer (British
Telecom), Simon Delord (France Telecom), Deborah Brungard (AT&T),
David McDysan (Verizon), Rahul Aggarwal (Juniper), Du Ke (ZTE),
Cagatay Buyukkoc (ZTE), and Stewart Bryant (Cisco).
9.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC3916] Xiao, X., Ed., McPherson, D., Ed., and P. Pate, Ed.,
"Requirements for Pseudo-Wire Emulation Edge-to-Edge
(PWE3)", RFC 3916, September 2004.
[RFC3985] Bryant, S., Ed., and P. Pate, Ed., "Pseudo Wire Emulation
Edge-to-Edge (PWE3) Architecture", RFC 3985, March 2005.
9.2. Informative References
[i610] Recommendation I.610 "B-ISDN operation and maintenance
principles and functions", February 1999.
[RFC5085] Nadeau, T., Ed., and C. Pignataro, Ed., "Pseudowire
Virtual Circuit Connectivity Verification (VCCV): A
Control Channel for Pseudowires", RFC 5085, December 2007.
[RFC4447] Martini, L., Ed., Rosen, E., El-Aawar, N., Smith, T., and
G. Heron, "Pseudowire Setup and Maintenance Using the
Label Distribution Protocol (LDP)", RFC 4447, April 2006.
[RFC4111] Fang, L., Ed., "Security Framework for Provider-
Provisioned Virtual Private Networks (PPVPNs)", RFC 4111,
[PWE3-OAM] Nadeau, T., Ed., Morrow, M., Ed., Busschbach, P., Ed.,
Alissaoui, M.,Ed., D. Allen, Ed., "Pseudo Wire (PW) OAM
Message Mapping", Work in Progress, March 2005.
[RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, RFC
2914, September 2000.
[RFC4364] Rosen, E. and Y. Rekhter, "BGP/MPLS IP Virtual Private
Networks (VPNs)", RFC 4364, February 2006.
[RFC4107] Bellovin, S. and R. Housley, "Guidelines for Cryptographic
Key Management", BCP 107, RFC 4107, June 2005.
117 West Street
Waltham, MA 02145
Alcatel-Lucent Telecom Ltd,
Cisco Systems, Inc.
9155 East Nichols Avenue, Suite 400
Englewood, CO, 80112
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