Internet DRAFT - draft-pan-pwe3-over-sub-ip

draft-pan-pwe3-over-sub-ip









Network Working Group                                        Ping Pan
Internet Draft                                   (Hammerhead Systems)
Expiration Date: December 2005                              July 2005




     Dry-Martini: Supporting Pseudo-wires in Sub-IP Access Networks


                     draft-pan-pwe3-over-sub-ip-01.txt


Status of this Memo

By submitting this Internet-Draft, each author represents that any
applicable patent or other IPR claims of which he or she is aware have
been or will be disclosed, and any of which he or she becomes aware will
be disclosed, in accordance with Section 6 of BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task
Force (IETF), its areas, and its working groups.  Note that other groups
may also distribute working documents as Internet-Drafts.

Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time.  It is inappropriate to use Internet-Drafts as reference material
or to cite them other than as ``work in progress.''

The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt

The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.


Copyright Notice

Copyright (C) The Internet Society (2005).  All Rights Reserved.
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.


Abstract

Several recent developments have significantly affected the carrier
access networks.  First, all large carrier backbones have migrated to
MPLS. Second, carriers are upgrading access circuits with less expensive
and high-speed Ethernet. Finally, the carriers will be offering advanced
data services over the access network infrastructure. Subsequently, the



Pan                                                             [Page 1]





Internet Draft                Dry-Martini                      July 2005


carriers have to face challenges in migration cost, data transport
efficiency and edge-to-edge user traffic management.

The Dry-Martini architecture is designed to help the carriers to
alleviate these challenges. It provides an approach to establish and
maintain pseudo-wires over any access network infrastructure, while
stripping off much of the IP/MPLS routing and signaling features that
are irrelevant in access network. As a result, all the existing
transport equipment, such as SONET/SDH MSPP, can provide MPLS pseudo-
wire functionality without much change to the existing platform.
Further, due its simplicity, this approach allows the new access
devices, such as PONĖs and Ethernet CPEĖs, to maintain low cost, while
being able to interface with MPLS networks.

This document assumes that the reader has at least some familiarity with
MPLS and pseudo-wire technologies.



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 RFC 2119.



2. Introduction

   The challenges in expanding carrier networks are mainly taking place
   at access. The driver behind the access network expansion is to bring
   high-speed links toward the end users so that the carriers can offer
   extensive data services.

   Subsequently, the carriers have to face a number of issues:

   First of all, how to deal with the existing installation?
   Traditionally, carrier access networks consist of DS1/3 and OC3/12
   circuits supporting services such as Frame Relay and ATM. The
   networks are constructed with either SONET/SDH rings, and/or leased
   lines. Most of the customer traffic is carried over circuit
   interfaces, despite the fact that IP data is becoming the
   predominating traffic type.

   Carrying packets over circuit networks cannot take the advantage of
   statistical multiplexing gains that has made packet networks very
   efficient. According to a recent report from AT&T [ATT], the
   multiplexing gain in data-aware access/metro networks is in the range



Pan                                                             [Page 2]





Internet Draft                Dry-Martini                      July 2005


   of 3-4.

   Second, how to manage and control the access networks? Most likely,
   the carriers would like to adapt a single, coordinated architecture
   in access, metro and backbone networks. ItĖs noteworthy that such
   architecture does not necessarily imply that all traffic flows will
   be packetized and individually routed as IP packets. It simply means
   that it is preferable to have a somewhat consistent control-plane
   throughout the network. Given the backbone has already migrated to
   IP/MPLS, naturally, the rest of the network should try to adapt to
   something similar.

   Finally, how to deal with the emerging Metro Ethernet? Access
   technology normally changes very 8-10 years. Most recently the
   carrier transport Ethernet technology has become the choice of access
   in metro network. When examining the previous network access
   technologies, we will notice that X.86, Frame Relay, ATM and Ethernet
   all share the same set of properties at time when they were deployed:

    (a) More bandwidth: this helps data statistical multiplexing

    (b) Robustness: failure detection and recovery are critical

    (c) Traffic assurance: user flows need to be shaped and policed

    (d) Cheap: this is the essential element for wide deployment

   Hence, this poses the question: is it possible to engineer the access
   network in such as a way that they donĖt have to be re-engineered
   very so often?

   One way to solve all the above issues is to replace the entire
   network equipment (including those in access and metro area) with
   MPLS routers, and interconnect them with big Ethernet pipes. The
   problem with this approach is in its willful ignorance of carrier
   economics. Unless it is green-field deployment, we believe that all
   carriers will have to focus on migration and operation cost as they
   expand data-friendly access network, and leverage the existing access
   infrastructure as much as possible.

   In this document, we will present the Dry-Martini architecture that
   provides a solution that is transparent to Layer-2 access
   technologies and can aggregate any type of data traffic from access
   into metro core.

   Before the description of the architecture, we first evaluate some of
   the important components, and gain a better understanding on what is
   and is not important in this area of the network.



Pan                                                             [Page 3]





Internet Draft                Dry-Martini                      July 2005


3. Components of the Architecture


   The figure below illustrates various relevant components in the
   architecture.



                   |                         |
                   |<-- label negotiation -->|
                +--------+             +----------+    +----------+
    User-1 -----| Access |             |  Metro   |    |   MPLS   |
                | Device |-------------|Aggregator|----| Backbone |-...
    User-2 -----|        |   sub-IP    |          |    |  Router  |
                +--------+ data trunks +----------+    +----------+
                   |                         |
                   |<----- Pseudo-wires ---->|

           Figure-1: A simplistic view of the carrier network




3.1. Access Devices and Data Trunks

   Traditionally, a typical Access Device can be a circuit aggregation
   switch (e.g. CPE), or a SONET/SDH switch (e.g. MSPP and ADM). The
   data trunks are normally DS1/DS3 circuits running Frame Relay and
   ATM, or OC-3/OC-12 circuits running ATM. At Access Device, each user
   flow is mapped to one of the circuits.

   In recent years, the carriers have begun to deploy high-speed
   Ethernet circuits and lambda (photonic) as data trunks to the
   customers. Access Devices have evolved as well. The PONs (Passive
   Optical Networks) in various flavor have been evaluated and deployed
   for high-speed network access.

   The links that interconnect the access devices and aggregators can be
   (and not limited to) lambda (photonic), SONET/SDH cross-connects,
   ATM/Frame Relay circuits, or Ethernet VLAN connections. For clarity,
   we denote these links as sub-IP data trunks, the networks that setup
   and maintain sub-IP data trunks, as sub-IP networks.

   It is important to realize that, it is carrierĖs interest to
   aggregate as many user flows into a sub-IP data trunk as the SLA
   would allow, in taking the advantage of statistical multiplexing
   gains.




Pan                                                             [Page 4]





Internet Draft                Dry-Martini                      July 2005


   To support wide deployment, the Access Device must be reliable,
   simple-to-manage and cheap. Normally, each access device connects to
   the metro backbone through two links for protection purpose.
   Extensive IP routing functionality at access devices is not always
   necessary.



3.2. Metro Aggregators

   Depending on metro network topology, each aggregator may process
   traffic from dozens to hundreds of access devices. Upon grooming user
   traffic together, the aggregators feed the traffic toward the MPLS
   core.

   Typically, each aggregator can be a combination of multiple systems.
   They are responsible for multiplexing and de-multiplexing DS3/1/0 or
   OC-n circuits, and switching data packets. As the metro network
   becomes more data-centric, the aggregators are expected to be more
   data friendly.

   Some of the key requirements in supporting data traffic on
   aggregators are per-user-flow QoS, data trunk OAM, and protection and
   recovery in event of both equipment and facility failures.



3.3. Packet Flow ID

   Each user flow can be uniquely identified with a packet flow ID.
   Depending on the transport technology, the flow ID can be (and not
   limited to) Frame Relay DLCI, ATM VPI/VCI, Ethernet VLAN (including
   VLAN tag via Q-in-Q), and MPLS Label.

   Using packet flow IDs enables the carriers to offer logical circuits
   to the customers, and, thereby, manage and control user traffic at
   per-flow granularity.



3.4. Pseudo-wire Encapsulation

   Aggregating multiple flows into a shared data trunk requires packet
   flow ID encapsulation. There have been a number of encapsulation
   methods for various technologies, such as Q-in-Q for Ethernet.

   Draft-martini, known for its original designer Luca Martini, has
   presented an encapsulation method that enables the aggregation of



Pan                                                             [Page 5]





Internet Draft                Dry-Martini                      July 2005


   different types of Layer-2 flows into a single trunk, while retain
   some of the important service-specific characteristics ([PWE3-ETHER],
   [MARTINI-ATM], [MARTINI-FR]). Essentially it is to encapsulate the
   packets belong to a user flow with a MPLS header and a Layer-2
   specific control word. The MPLS label [RFC3032] in the header makes
   the flow unique throughout the network. Upon entering the network,
   each encapsulated flow is referred as a ŸPseudo-wire÷.

   Originally, draft-martini was designed to help the carriers to
   aggregate Layer-2 traffic over a common MPLS/IP backbone via MPLS
   Label Edge Routers (LERĖs). Each Layer-2 flow is mapped to a Pseudo-
   wire. The setup and maintenance of each Pseudo-wire involve two
   provider edge nodes (PEĖs), which exchange connection and
   encapsulation label information through targeted LDP [PWE3-CTRL].

   Upon closer evaluation, we note the following: one of the most
   important benefits in Pseudo-wire is that the carrier can handle and
   process user traffic at per-flow granularity independent of the
   underlying networking technology. On the contrary, encapsulation
   mechanisms, such as Ethernet Q-in-Q, will only operate in a specific
   Layer-2 environment. We will elaborate this point later on.



3.5. Label Negotiation

   Pseudo-wires operate between two network nodes. Each Pseudo-wire
   consists of two unidirectional paths, one in each direction. Each
   path can be uniquely identified by the triple <node-1, node-2,
   encapsulation-label>. Prior to data transmission, two nodes need to
   know the value of the encapsulation-label for each user flow.

   As mentioned above, the most common label negotiation mechanism is
   target LDP [LDP, PWE3-CTRL], where two edge nodes can initiate a LDP
   session and exchange label binding information to setup the Pseudo-
   wires.

   The actual Pseudo-wire setup sequence is very simple. However, the
   overhead in processing LDP protocol itself, such as LDP session and
   adjacencies management and peer discovery can be quite elaborate.
   This is because the original LDP specification was designed to
   establish MPLS LSPĖs among routers in the backbone environment.
   Hence, applying the same label negotiation mechanism in access
   network may be an over-kill. We will discuss this issue later on.







Pan                                                             [Page 6]





Internet Draft                Dry-Martini                      July 2005


3.6. Traffic Policing and Assurance

   Most of the data applications do not require stringent QoS. In
   todayĖs backbone networks, the carriers over-provision the network
   links, and rely on DiffServ to overcome temporary traffic congestion.
   Thus, per-flow shaping and rate limiting does not apply in enterprise
   and backbone networks. However, we should not ignore QoS guarantees
   (or SLA) as an essential part of carrier service offering.

   As the access network begins to expand, the mixture of existing low-
   speed circuit infrastructure and high-speed Ethernet links will cause
   network resource heterogeneously distributed. As carriers continue to
   lease service lines to enterprise customers, and offer QoS-sensitive
   data applications (e.g. voice) to consumer users, supporting per-
   user-flow QoS becomes increasingly important.



3.7. Demarcation Points and Pseudo-wire Switching

   The concept of UNI (User-Network Interface) has been foreign to the
   Internet, where the entire network supposes to be distributed and
   open. From user application perspective, this is true: the end-users
   should never observe the existence of any UNI interface, as
   elaborately defined in ATM.

   Nevertheless, in access/metro area, there remain demarcation points
   where traffic from one segment of the network is transferred to
   another accordingly to a set of rules between carriers. For instance,
   there likely exists a demarcation point between metro network
   (aggregators) and the backbone (MPLS/IP routers). As such, depending
   on carrierĖs deployment scenario, the Pseudo-wires coming from access
   networks need to be policed before switching toward the backbone.

   Another demarcation point where subsequent Pseudo-wire switching is
   useful is between two metro networks. Depending on the bilateral (and
   multilateral) agreement, user flows in the form of Pseudo-wires will
   be handed off (or switched, stitched) from one network to another.













Pan                                                             [Page 7]





Internet Draft                Dry-Martini                      July 2005


4. Dry-Martini Architecture

   The Dry-Martini Architecture is based on draft-martini as the
   encapsulation method for aggregating user flows into carrier
   networks, and strip off much of the IP/MPLS routing and signaling
   features that are irrelevant in access networks. Because we have
   relaxed and simplified much of the constraints in the original
   design, we refer this architecture as Ÿdry-martini÷.

   Our rational is as follows: the operation of Pseudo-wires involves
   two endpoints of the connection, and is almost independent from the
   operation taking place within the underlying network. Hence, we argue
   that the pseudo-wire concept is applicable in all networks. (Note
   that much of the concept of supporting Pseudo-wires over any packet
   networks (PSN) has been assumed in the IETF PWE3 architecture [PWE3-
   ARCH].)

   Further, pseudo-wire encapsulation is transparent to all the existing
   Layer-2 technologies, and, perhaps, to the access technologies that
   will get deployed in the future (such as WiMax). Therefore, if we
   always handle user flows at Pseudo-wire layer, it will provide
   carriers with a uniformed and consistent method to aggregate, police
   and manage all types of data flows. This will subsequently simplify
   carrier service migration and integration.

   Another advantage in Pseudo-wires is in its flexibility to map user
   applications into flows.

   For example, today, the carriers can packetize voice streams into
   Pseudo-wires [SATOP, CESoPSN] and transport them over the backbone.
   Voice service requirements can be retained with some basic QoS
   treatment (such as clocking recovery). This implies that, if the
   carriers can manage data flows as Pseudo-wires with QoS, redundancy
   and OAM features, then any stream traffic (such as video and voice)
   sent over Pseudo-wires will always get appropriate edge-to-edge SLA
   guarantees.

   In essence, Pseudo-wire can be the convergence layer for transporting
   data flows over any network. Figure 2 shows the Pseudo-wire in
   relation to user applications and underlying transport network.











Pan                                                             [Page 8]





Internet Draft                Dry-Martini                      July 2005




       +-------------------+       user         +-------------------+
       |     Payload       |   applications     |     Payload       |
       | (IP, L2 data, ... |<==================>| (IP, L2 data, ... |
       | packetized voice) |                    | packetized voice) |
       +-------------------+                    +-------------------+
       |                   |     convergence    |                   |
       |                   |        layer       |                   |
       |   Pseudo-wires    |<==================>|    Pseudo Wires   |
       |                   |                    |                   |
       +-------------------+                    +-------------------+
       |     Layer 2       |     data service   |       Layer 2     |
       |(Ethernet, ATM...) |<==================>|(Ethernet, ATM...) |
       |                   |                    |                   |
       +-------------------+                    +-------------------+
       |     Layer 1       | physical transport |       Layer 1     |
       | (optical, TDM...) |<==================>| (optical, TDM...) |
       |                   |                    |                   |
       +-------------------+                    +-------------------+


              Figure 2: The logical layering model applied
                          in Dry Martini architecture


   Within the Dry-Martini architecture, the carriers can operate
   Pseudo-wires over any sub-IP networks:

   Application 1: The carriers may create Pseudo-wires from SONET/SDH
   access devices (such as MSPPĖs) directly, and aggregate user Ethernet
   traffic over the existing metro infrastructure. TodayĖs typical EoS
   (Ethernet-over-SONET) solution is to use VCAT and GFP, and map
   Ethernet physical port to a Virtual Concatenation Group (VCG). With
   the Dry-Martini architecture, the carriers can map multiple Ethernet
   flows into a single VCG. This can improve access bandwidth
   utilization significantly.

   Application 2: The carriers may choose to create Pseudo-wires and
   aggregate data packets over the existing leased ATM or Frame Relay
   circuits. Once again, improving link utilization (a.k.a. statistical
   multiplexing gains) may be an important economical factor here.

   Application 3: The carriers can always aggregate multiple user flows
   into a single wavelength off the PONĖs, and process the flows at
   aggregators.

   In all the applications, carriers use the method of their choice to



Pan                                                             [Page 9]





Internet Draft                Dry-Martini                      July 2005


   manage the sub-IP data trunks: GMPLS for optical networks, PNNI/SPVC
   for ATM infrastructure, Ethernet signaling for metro Ethernet, or,
   simply, static configuration for SONET/SDH connection provisioning.
   The key is that, as long as there is an operational sub-IP data trunk
   between two network nodes, the carrier may establish Pseudo-wires to
   aggregate data flows over it.

   In the architecture, the access devices do not need to support much
   of the IP functionality, such as per-packet IP routing, and routing
   protocols. All data flows are handled as circuit-like Pseudo-wires.
   Given the access devices have only a couple of connections to the
   metro backbone, the use of IP routing would be very limited. It is
   the aggregators that need to be able to interface with both access
   networks and the MPLS/IP network, and aggregate and switch user
   traffic in between.

   In the remaining of the document, we will outline the data-plane and
   control-plane issues in supporting the Dry-Martini architecture, and
   articulate some of the important features that are actually needed in
   this area of the network.



5. Data-Plane Operation in the Dry-Martini architecture

   As an example of the operation, we consider the network setup shown
   in Figure-1. Suppose that there are N user flows arriving at the
   Access Device. The user flows are provisioned prior to the actual
   data transmission. A typical user flow may be a privately leased line
   for an enterprise customer, or a high-speed Ethernet connection to a
   residential location. Thus, we assume that all the user flows have a
   relatively long duration, and each flow may have some QoS parameters
   (such as average rate) associate with it.

   Upon the reception of a packet from a user, the access device will
   first search for the corresponding flow information (such as the VLAN
   tag) from its local cached database. If a match is found, it will run
   a simple CAC (Call Admission Control) algorithm to ensure QoS
   compliance and encapsulate the packet with a new packet flow ID, and
   then transmit the packet toward the aggregator.

   The packet flow ID is negotiated with the aggregator ahead of time.
   Multiple user flows can share a common data trunk with different flow
   IDĖs.

   At the aggregator, the packet flow ID will be examined and stripped
   off. The packets will then be forwarded toward the core.




Pan                                                            [Page 10]





Internet Draft                Dry-Martini                      July 2005


   The packet forwarding sequence from the aggregator to the access
   device would be same.

   All packet encapsulation should be the same as the ones defined in
   draft-martini. However, we do not mandate the MPLS label as the only
   packet flow ID. Depending on the flexibility of the access device
   itself, the packet flow ID can be something different, for instance,
   Ethernet VLAN tag.

   Ethernet VLAN can be used as the Pseudo-wire flow ID between access
   devices and aggregators. This is reasonable for the following
   reasons: first, given the network size between access devices and
   aggregators, VLAN scaling (that is, 4000 VALN tags per interface)
   should not a real problem. Second, some of the access devices may not
   have the ability to support various MPLS label manipulation (push,
   pop and stack) operations.

   In summary, the data-plane requirement on ant access device is very
   trivial. No per-packet routing lookup is required. However, the
   access device needs to be able to police user traffic on per-flow
   basis.

   The procedure on the aggregator is similar, expect that the
   aggregators interface with the core, and may need to exercise more
   extensive packet processing functions with the core routers.



6. Control-Plane in the Dry-Martini architecture

   As mentioned before, label negotiation takes place between the access
   devices and the aggregators.

   In practice, there are two general approaches in setting up Pseudo-
   wire labels: in-band and out-band. The in-band approach is to
   exchange control messages over the sub-IP data trunks. The out-band
   approach is to setup labels through an external management system.

   There are a number of ways to exchange IP control messages (e.g., LDP
   messages) between the edge nodes. One approach is to route the
   messages through the underlying sub-IP network. For example, in
   SONET/SDH networks, the control messages may utilize DCC channels to
   communication with each other. However, this would require every node
   in the sub-IP network to be IP-capable, which may be not practical in
   many of the operational networks.

   We propose to "tunnel" all control packets through the sub-IP data
   trunks as regular data payload from the edge. The advantage here is



Pan                                                            [Page 11]





Internet Draft                Dry-Martini                      July 2005


   that the exchange of control messages will not disturb the operation
   in the underlying sub-IP network.

   For the user packets to be encapsulated with a MPLS header, we
   require control packets to be encapsulated with "IP4 Explicit NULL
   Label" [RFC2032]. At the destination, the label is popped, and the
   control packets are delivered to the control plane for further
   processing.

   For the user packets to be encapsulated with a VLAN tag, we propose
   to use a special VLAN value for control message delivery. Once again,
   at the destination, the messages are picked up for further
   examination.



6.1. Option 1: Target LDP

   The conventional method is to run target LDP in-band between access
   device and aggregator. For clarity, we repeat the procedure described
   in [PWE3-CTRL] in the context of setting up Pseudo-wires in sub-IP
   networks.

   Each Pseudo-wire consists of two unidirectional paths, one in each
   direction. The access device and the aggregator are the two edge
   nodes. Each edge initiates the setup of the path on behalf of ingress
   Layer-2 traffic. Each path is uniquely identified by the triple <PE-
   1, PE-2, VCID>, where the VCID is a 32-bit quantity that must be
   unique in the context of a single LDP session between two edges. For
   a given Pseudo-wire, the same VCID must be used when setting up both
   paths.

   In this case, the access device needs to implement target LDP to
   communicate with the aggregator.



6.2. Option 2: Lightweight Signaling

   In certain scenarios, using target LDP for Pseudo-wire label
   negotiation is questionable. LetĖs first evaluate the tradeoffs in
   supporting target LDP in this part of the network.

   As mentioned before, much of the LDP functionality is irrelevant in
   setting up Pseudo-wires. Its built-in discovery and session and
   adjacencies management algorithms are originally designed to operate
   in a distributed networking environment and interface among routers.




Pan                                                            [Page 12]





Internet Draft                Dry-Martini                      July 2005


   However, in access networking environment, the network configuration
   is most likely a simple spoke topology. In most cases, the access
   devices and the aggregators are one-hop away.

   From the operation point of view, the carriers would like to control
   the label allocation and distribution at the aggregators, rather than
   from some customer-location access devices. The actual Pseudo-wire
   negotiation signaling should be more client-server style, instead of
   peer-to-peer as in LDP.

   Further, letĖs examine the development and performance cost. One of
   the primary requirements in this part of the network is low-cost.
   Supporting MPLS signaling and IP routing protocols will no doubt
   require more expensive components in developing the access devices.
   Typically, each metro aggregator may interface with hundreds of
   access devices. Supporting target LDP implies that each aggregator
   may need to maintain hundreds of TCP and LDP sessions at control-
   plane. This adds unnecessary performance overhead to the aggregators.

   We believe that there is a need for developing a light-weight
   Pseudo-wire negotiation signaling protocol for access networks. Some
   of the key elements of the protocols are:

   (a) Client-server signaling: If we look at the operation of Frame
   Relay LMI, in practice, the DTE-DCE relationship is nearly identical
   to that of access devices and metro aggregators.

   (b) Light-weight: The complexity of the protocol should be similar to
   that of ICMP. In other words, every access device should be able to
   support it without much cost associated with it.

   (c) In-band: This will increase the automation capability during
   Pseudo-wire setup. Further, in-band protocols can always help the
   network nodes in failure detection.

   The detail design of the lightweight protocol is beyond the scope of
   this architecture document. We will provide its design elsewhere.

   In summary, using a lightweight protocol would enable the access
   devices to setup pseudo-wires without dealing with IP routing and
   full-stack MPLS signaling. When designed correctly, the aggregators
   would have an efficient control-plane interface with the access
   devices.








Pan                                                            [Page 13]





Internet Draft                Dry-Martini                      July 2005


6.3. Option 3: Pseudo-wire Proxy

   This method is to deploy Ÿproxies÷ to manage Pseudo-wire allocation
   and distribution. Figure 3 shows its configuration. The Pseudo-wire
   proxy is a logical entity that can operate in carrierĖs NOC, or on
   metro aggregators.



                       +-----------------+
                   +---|Pseudo-wire Proxy|---+
                   |   +-----------------+   |
                   |                         |
                +--------+             +----------+      +----------+
    User-1 -----| Access |             |  Metro   |      |   MPLS   |
                | Device |-------------|Aggregator|------| Backbone |
    User-2 -----|        |   sub-IP    |          |      |  Router  |
                +--------+ data trunks +----------+      +----------+
                   |                         |
                   |<--------- OAM --------->|

           Figure-3: Pseudo-wire management with a Pseudo-Wire Proxy


   The operation sequence can be as follows: a customer negotiates with
   the carrier on his network access service, which may include an
   extensive set of business and technical conditions. During the
   negotiation, the carrier knows where the customerĖs traffic will
   enter the network, and the access devices, the aggregators and the
   data trunks to be used. From the proxy, the carrier assigns the
   Pseudo-wire labels to both the access device and the aggregator
   through the management interface (e.g. SNMP).

   This type of out-band, static Pseudo-wire configuration is simple to
   implement. However, without some type of direct communication between
   the access devices and the aggregators, any failure in the data trunk
   will not be detected, or too late to be notified. Hence, we recommend
   enabling the Pseudo-wire OAM functionality when operate in this mode.













Pan                                                            [Page 14]





Internet Draft                Dry-Martini                      July 2005


7. Carrier Deployment Considerations


7.1. Pseudo-wire Switching

   As shown in Figure 1, the aggregators interface with the access
   devices via pseudo-wires, and can interface with the core routers
   with MPLS or Pseudo-wires. By switching pseudo-wires between the
   access networks and the metro core, the carriers would have the
   ability to control the user flows edge-to-edge.

   Much of the work in Pseudo-wire switching is taking place in IETF at
   the moment. We will not expand the description any further at this
   point.



8. Security Considerations

   This document extends the use of PWE3 to sub-IP networks. It has the
   same security requirements as in PWE3.



9. Contributors

   Tad Hofmeister, Tedi Tedijianto and Calcin Leung have contributed to
   the original dry-martini ideas back in the fall of 2002. Much of the
   ideas and concepts have been thoroughly discussed and validated with
   a number of my colleagues in Hammerhead Systems and operation and
   architecture folks in various carriers.



10. Normative Reference

   [ATT] T. Afferton, et al, "Packet Aware Transport for Metro
   Networks", IEEE Network Magazine, April 2004.

   [RFC3032] E. Rosen, et al, "MPLS Label Stack Encoding".

   [PW-CTRL] L. Martini, et al, "Pseudo-wire Setup and Maintenance using
   LDP", draft-ietf-pwe3-control-protocol-14.txt

   [LDP] L. Andersson, et al, "LDP Specification", draft-ietf-mpls-
   rfc3036bis-00.txt

   [PWE3-ARCH] S. Bryant, P. Pate, "PWE3 Architecture", draft-ietf-



Pan                                                            [Page 15]





Internet Draft                Dry-Martini                      July 2005


   pwe3-arch

   [PWE3-ETHER] L. Martini, et al, "Encapsulation Methods for Transport
   of Ethernet Frames over IP/MPLS Networks", draft-ietf-pwe3-ethernet-
   encap

   [MARTINI-ATM] L. Martini, et al, "Encapsulation Methods for Transport
   of ATM Cells/Frame over IP and MPLS Networks", draft-martini-atm-
   encap-mpls

   [MARTINI-FR] C. Kawa, et al, "Frame Relay Encapsulation over Pseudo-
   wires", draft-martini-frame-encap-mpls

   [CESoPSN] A. Vainshtein, et al, ?Structure-aware TDM Circuit
   Emulation Service over Packet Switched Network?, IETF Draft



11. Informative References

   [PWE3-TRANSPORT] L. Martini, et al, "Transport of Layer 2 Frames over
   MPLS", draft-martini-l2circuit-trans-mpls

   [SAToP] A. Vainshtein, Y. Stein, ?Structure-Agnostic TDM over
   Packet?, IETF Draft

   [TDMoIP] Y. Stein, et al, ?TDM over IP?, IETF Draft

   [CEP] A. Malis, ?SONET/SDH Circuit Emulation over Packet?, IETF Draft


12. Author Information


   Ping Pan
   Hammerhead Systems
   640 Clyde Court
   Mountain View, CA 94043
   e-mail: ppan@hammerheadsystems.com



   Intellectual Property Statement

   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



Pan                                                            [Page 16]





Internet Draft                Dry-Martini                      July 2005


   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

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

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

   By submitting this Internet-Draft, I certify that any applicable
   patent or other IPR claims of which I am aware have been disclosed,
   or will be disclosed, and any of which I become aware will be
   disclosed, in accordance with RFC 3668.


   Full Copyright Statement

   Copyright (C) The Internet Society (2005).  This document is subject
   to the rights, licenses and restrictions contained in BCP 78, and
   except as set forth therein, the authors retain all their rights.

   This document and the information contained herein are provided on an
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
   ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
   INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
   INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
   WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.


   Acknowledgment

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








Pan                                                            [Page 17]