Application of
Ethernet Pseudowires to MPLS Transport NetworksCisco Systems250, Longwater, Green Park,ReadingRG2 6GB, UKUKstbryant@cisco.comCisco SystemsGlatt-comCH-8301 GlattzentrumSwitzerlandmmorrow@cisco.comCisco Systems1414 Massachusetts AveBoxboroughMA01719swallow@cisco.comJuniper Networks,1194 N. Mathilda AveSunnyvaleCA 94089BTtom.nadeau@bt.comBTneil.2.harrison@bt.comBT208 Callisto House, Adastral ParkIpswichSuffolkIP5 3REUKbenjamin.niven-jenkins@bt.com
Internet Area
Network Working GroupSampleDraftA requirement has been identified by the operator community for the
transparent carriage of the MPLS(-TP) network of one party over the
MPLS(-TP) network of another party. This document describes a method of
satisfying this need using the existing PWE3 Ethernet pseudowire
standard RFC4448.The operator community has identified the need for the transparent
carriage of the MPLS(-TP) network of one party over the MPLS(-TP)
network of another party . This document describes
one mechanism to satisfy this requirement using existing IETF standards
such as PWE3 Ethernet pseudowire standard
. The mechanism described here fulfills the MPLS-TP requirements for
transparent carriage (MPLS-TP requirements 20 & 21) of the Ethernet
data plane.The key purpose of this document is to demonstrate that there is an
existing IETF mechanism with known implementations that satisfies the
requirements posed by the operator community. It is recognised that it
is possible to design a more efficient method of satisfying the
requirements, and the IETF anticipates that improved solutions will be
proposed in the future.Much of the notation used in this document is defined in to which the reader is referred for
definitions.The architecture required for this mechanism is illustrated in Figure
1 below.An 802.3 (Ethernet) circuit is established between CE1 and CE2. This
circuit may be used for the concurrent transport of MPLS packets as well
as IPv4 and IPv6 packets. The MPLS packets may carry IPv4, IPV6, or
Pseudowire payloads, and Penultimate-Hop-Popping (PHP) may be used. For
clarity these paths are labeled as the client in Figure 1.An Ethernet pseudowire (PW) is provisioned between PE1 and PE2 and
used to carry the Ethernet from PE1 to PE2. The Ethernet PW is carried
over an MPLS packet switched Netwok (PSN), but this PSN must not be
configured with PHP. For clarity this Ethernet PW and the MPLS PSN are
labeled as the server in Figure 1. In the remainder of this draft call
the server network a transport network.The PWE3 encapsulation used by this specification to satisfy the
transport requirement is Ethernet . This
is used in "raw" mode.The Control Word must be used. The Sequence number must be zero.The use of the Pseudowire Setup and Maintenance Label Distribution
Protocol is not required by the profile
of the PWE3 Ethernet pseudowire functionality defined in this
document.The Pseudowire Label is statically provisioned.Within a connection, traffic units sent from the single source are
constrained to stay within the connection under defect-free conditions.
During misconnected defects, a connection can no longer be assumed to be
constrained and traffic units (and by implication also OAM packets) can
'leak' uni-directionally outside a connection. Therefore during a
misconnected state, it is not possible to rely on OAM which relies on a
request/response mechanism ; and, for this reason such OAM should be
treated with caution if used for diagnostic purposes.Further, when implementing an Equal Cost Multi-path (ECMP) function
with MPLS, use of the label stack as the path selector such that the OAM
and data are not in a co-path as any failure in the data path will note
be reflected in the OAM path. Therefore, an OAM that is carried within
the data-path below the PW label such as Virtual Circuit Connectivity
Verification (VCCV) is NOT vulnerable to the above failure mode. For
these reasons the OAM mechanism is , using
Bidirectional Forwarding Detection (BDF) for connection verification (CV). The
method of using Bidirectional Forwarding Detection (BFD) as a CV method
in VCCV is described in [I-D.draft-ietf-pwe3-vccv-bfd] . One of the VCCV
profiles described in Section 3.1 or Section 3.2 must be used. Once a
VCCV control channel is provisioned, and the operational status of the
PW is UP, no other profile SHOULD be used until such time as the PW's
operational status is set to DOWN.When PE1 and PE1 are not IP capable or have not been configured
with IP addresses, the following VCCV mechanism SHOULD be used.The connection verification method used by VCCV is BFD with
diagnostics as defined in [I-D.draft-ietf-pwe3-vccv-bfd]. specifies that the first nibble is
set to 0x1 to indicate a channel associated with a pseudowire .The Version and the Reserved fields are set to zero, and the
Channel Type is set to [TBD] to indicate that the payload carried is
BFD without IP/UDP headers, as is defined in
[I-D.draft-ietf-pwe3-vccv-bfd].When PE1 and PE1 are IP capable and have been configured with IP
addresses, the following VCCV mechanism may be used.The connection verification method used by VCCV is BFD with
diagnostics as defined in [I-D.draft-ietf-pwe3-vccv-bfd]. specifies that the first nibble is
set to 0x1 to indicate a channel associated with a pseudowire .The Version and the Reserved fields are set to 0, and the Channel
Type is set to 0x21 for IPv4 and 0x56 for IPv6 payloads .The architecture of MPLS enabled networks is described in . This section describes a subset of the
functionality of the MPLS enabled PSN. There are two cases that need to
be considered: The case where external configuration is used.The case where a control plane is available.Where the use of a control plane is desired this may be based on
Generalized Multi-Protocol Label Switching (GMPLS) The use of external provisioning is not precluded from being
supported by the current MPLS specifications. It is however expicitly
described in this specification to addess the requirements specified
by the ITU to
address the needs in a transport environment.The MPLS encapsulation is specified in . All MPLS labels used in the server layer (Figure 1) must be
statically provisioned. Labels may be selected from either the
per-platform or the per-interface label space.All transport Label Switched Paths (LSPs) utilized by the PWs
described in section 2 must support both unidirectional and
bi-directional point-to-point connections.The transport LSPs SHOULD support unidirectional
point-to-multipoint connections.The forward and backward directions of a bi-directional connection
should follow a symmetrically routed (reciprocal) LSP in the server
network.Equal cost multi-path (ECMP) load balancing must not be configured
on the transport LSPs utilized by the PWs described in sections 2.The merging of label switched paths is prohibited and must not be
configured for the transport LSPs utilized by the PWs described in
section 2.Penultimate hop popping by the transport label switched routers
(LSRs) must be disabled on transport LSPs.Both EXP-Inferred-PSC LSPs (E-LSP) and Label-Only-Inferred-PSC LSPs
(L-LSP) must be supported as defined in .For the MPLS EXP field only the pipe and short-pipe models are
supported.In this section we describe the control plane configuration
when “RSVP-TE: Extensions to RSVP
for LSP Tunnels” or the bi-directional support in GMPLS “Generalized Multi-Protocol Label
Switching (GMPLS) Signaling Functional Description" and “Generalized Multi-Protocol Label
Switching (GMPLS) Signaling Resource ReserVation Protocol-Traffic
Engineering (RSVP-TE) Extensions” are used to configure the
transport MPLS PSN. When these protocols are used to provide the
control plane the following are automatically provided:There is no label merging unless it is deliberately enabled to
support Fast Re-route (FRR) .A single path is provided end-to-end (there is no ECMP).Label switched paths may be unidirectional or bidirectional as
required.Additionally the following configurations restrictions required to
support external configuration must be applied:Penultimate hop popping by the LSRs must be disabled on LSPs
providing PWE3 transport network functionality .Both E-LSP and L-LSP must be supported as defined in .The MPLS EXP field is supported
according to for only when the pipe
and short-pipe models are utilized.This draft describes a method of using the existing PWE3 Ethernet
pseudowire to solve a particular network
application. The congestion considerations associated with that
pseudowire and all subsequent work on congestion considerations
regarding Ethernet pseudowires is applicable to this draft.This draft is a description of the use of existing IETF proposed
standards to solve a network problem, and raises no new security
issues.The PWE3 security considerations are described in and the Ethernet pseudowire security
considerations of.The Ethernet pseudowire is transported on an MPLS PSN; therefore, the
security of the pseudowire itself will only be as good as the security
of the MPLS PSN. The server MPLS PSN can be secured by various methods,
as described in.The use of static configuration exposes an MPLS PSN to a different
set of security risks to those found in a PSN using dynamic routing. If
a path is missconfigured in a staticly configued network the result can
be a persistent black hole, or much worst, a persistent forwarding loop.
On the otherhand most of the distributed components are less complex.
This is however offset by the need to provide failover and redundancy in
the management and configuration system and the communications paths
between those central systems and the LSRs.Security achieved by access control of media access control (MAC)
addresses , and the security of the client layers is out of the scope of
this document.There are no IANA actions required by this draft.The authors wish to thank Matthew Bocci, John Drake, Adrian Farrel,
Andy Malis, and Yaakov Stein for their review and proposed enhancements
to the text.