Tutorial on a critical convergence technology: * What they are * How they work * Applications

July 7, 2005

19 Min Read
Pseudowires

Pseudowire solutions began quietly as a way to transport legacy services over operators’ Multiprotocol Label Switching (MPLS) cores. Today, however, pseudowires are proliferating and increasing in importance, despite winning the Light Reading award for the most ugly and opaque marketing term in telecom history.

So what are pseudowires all about? Are they really:

  • A true end-to-end solution for operators?

  • A key solution for convergence, beyond Layer 2 services?

  • Transforming the access network by tightly integrating it with the core?

  • Providing a platform for new services (such as Virtual Private LAN Service), and not just for the transport of legacy services?

Because "pseudowire" is a pretty new term commercially (although it has been around in the IETF for a couple of years or so) there’s quite a lot confusion over what they can be used for and what their real significance is. This Light Reading report explains the basics and shows how pseudowires are key to such applications as:

  • Metro Ethernet over MPLS

  • Multiservice access

  • MSO commercial services

  • Multiservice backhaul

  • Mobile applications

  • FTTP backhaul

Here’s a hyperlinked contents list:

  • Pseudowire Basics
    Anything-over-packets means real convergence

  • Pseudowire Evolution
    It's more and more deployable

  • Pseudowire Applications I
    Lots and lots of useful apps...

  • Pseudowire Applications II
    ...and not just for plain vanilla carriers

Webinar

This report was previewed in a Webinar moderated by Scott Clavenna, Chief Analyst, Heavy Reading. Sponsored by Axerra Networks Inc., Mangrove Systems Inc., and Overture Networks Inc., it may be viewed free of charge in our Webinar archives by clicking here.

Background Reading on Light Reading

  • Column: Supercomm Review

  • News Wire Feed: Patapsco Offers ISDN Over IP Tunneling

  • Column: How Far Can MPLS Go?

  • Column: How About a Dry Martini?

Related Light Reading Webinar archives

Related Heavy Reading reports

— Tim Hills is a freelance telecommunications writer and journalist. He's a regular author of Light Reading reports.

According to the Internet Engineering Task Force (IETF), pseudowire emulation edge-to-edge (PWE3) is a mechanism that emulates the essential attributes of a service such as ATM, Frame Relay, or Ethernet over a packet-switched network (PSN). The basic idea is that there is a Layer 3 network over which an operator wants to transport legacy services, including Layer 2 services.

This makes pseudowires a powerful tool for convergence now that operators worldwide have built big IP core networks and are extending MPLS towards the edge of those networks. They can transport legacy services that are already generating revenues and with which customers are already familiar, but can take advantage of the high speed and wide connectivity of their new and scaleable IP/MPLS networks to lower the cost of legacy services and to extend them into new markets. They can also support new services to provide new sources of revenue.

Figure 1 shows the basic concept of a pseudowire. Pseudowires (PW) are defined to run over IP networks, Layer Two Tunneling Protocol (L2TP) networks, and also, and more commonly, MPLS networks. These networks provide the packet "cloud" through which connection-oriented tunnels are formed to support pseudowires. For the common MPLS case, two unidirectional, inner-tunnel, label-switched paths (LSPs) are contained within unidirectional, outer-tunnel LSPs (which act as traffic-engineering tunnels) and create a bidirectional connection between provider edge (PE) routers.

75315_1.gifThe inner LSPs form the pseudowires by using an interworking function (IWF) – currently residing at the PE, although it can easily and rapidly migrate to the customer equipment (CE) – that encapsulates the CE data transmission format (such as Frame Relay, ATM, or Ethernet) from the attachment circuit into the IETF-defined PWE3 format of the pseudowire. At the far end of the pseudowire, the data is unencapsulated and handed over to the destination CE.

Currently, like-to-like attachment circuits over point-to-point PWs are the main implementations, but future implementations will offer service interworking among attachment circuits, multipoint service architectures, dynamic signaling end-to-end, and operations, administration, and maintenance (OA&M) end-to-end.

Pseudowires for Convergence

The basic point of pseudowires is that they decouple services – protocols and applications – from the underlying facilities carrying them. Provided a network can support IP or IP/MPLS, it can support pseudowires and thus any other service by running it over pseudowires. This makes pseudowires a natural technology for building fully converged networks, especially given the general move of operators and carriers worldwide to invest in large IP/MPLS core networks. So an operator could offer a T1-service pseudowire, for example, anywhere in its network where IP with sufficient bandwidth is available, and also transport that T1 pseudowire across its IP backbone/core.

Pseudowires can thus replace today’s mix of separate Frame Relay, ATM, and packet-switched networks with a single converged network, and with considerable savings in complexity and costs.

“The true goals of convergence with pseudowires is to reduce costs: It’s one network to provision; it’s one network to maintain,” says Chip Redden, Senior Director, Marketing & Product Management, Overture Networks. “It’s also the ability to offer new services. We hear people say they are worried about revenue cannibalization, but they should really be worried about revenue migration. Pseudowires also give you the ability to carry legacy services that you are already making money from. So with pseudowires you can build a new network, build new services, and yet continue to offer the old ones at the same time.”

Pseudowires are an evolving technology. Standards are now moving outside the IETF to the International Telecommunication Union (ITU), the Metro Ethernet Forum (MEF), and the MPLS/Frame Relay Alliance, for example. These latter organizations are building services and implementation agreements.

The technology began with the well known Martini Draft VPNs as a legacy Layer 2 transport over MPLS or L2TP IP networks. (See Virtual Private Networks, Luca Martini, Level 3, Luca Martini Leaves Level 3, Luca's First Day at Cisco , and How About a Dry Martini?.) This proved to be a useful technology, and a lot of Ethernet and some Frame Relay and ATM have been deployed this way as transport solutions. Now there is the beginning of a move towards multipoint architectures such as VPLS, and also improvements to such Layer 2 architectures with Layer 2 service interworking. As a result, a single multipoint VPN connection can now have many different Layer 2 attachment circuits that have true service interworking. So a network could have a mix of Frame Relay or Ethernet connections, yet be interworked within the cloud of pseudowires.

Pseudowires are also beginning to move closer to the customer. The first transport-oriented use of pseudowires really leveraged the MPLS core, so the pseudowires did not begin until the edge of the core network. Now, however, access and metro devices are appearing that can perform the pseudowire encapsulation and, through recently developed dynamic-signaling pseudowire OA&M, allow end-to-end pseudowires to be created from customer location to customer location.

A key point is that it is not a requirement to have MPLS out at the customer location. Pseudowires can run over IP directly, including L2TPE3, and, with the work currently underway in the Metro Ethernet Forum, they can run over Ethernet directly as well. This makes the low-cost ($2,000 and below) CPE devices needed for mass deployment perfectly feasible.

Developments like this take pseudowires towards the mass deployment scale, where pseudowire solutions are not just associated with converging the network core, but really converge in the access network as well – to create convergent end-to-end infrastructures. Once further refinements – such as pseudowire switching, user–to-network and network–to-network interfaces (UNI/NNI), and signaling interworking – reach mass deployment towards 2008, pseudowires will fully enter the telecom mainstream.

Figure 2 shows there are now working standard pseudowires that enable a broad range of legacy and emerging services over packet networks, including T1/E1 services (both data and voice), Frame Relay, and ATM. Pseudowires can extend these services from existing Layer 2 service networks, doing simple backhaul known as port mode, and can provide shaping, policing, and service-level assurance, too.

75315_2.gifCapabilities include:

  • T1/E1 private-line data: Dynamic bandwidth sharing and support for SNA, X.25, encryption, and proprietary data protocols

  • T1/E1 voice: No change in PBX and Class 5 voice switching and signaling; dynamic-CES makes bandwidth available when no traffic is present

  • Frame Relay: True Frame Relay UNI or NNI per Frame Relay port

  • E-Line and E-LAN services

Dynamic Provisioning of Pseudowires

Connection-oriented services such as pseudowires over an MPLS backbone must be provisioned. Static provisioning works for small networks, but becomes much more difficult and high-maintenance in large provider networks, and so some form of dynamic provisioning is needed. There is an echo here of the earlier situation with ATM, also connection-oriented, which many carriers and operators found complex, difficult, and error-prone to provision, despite the arrival of ATM signaling.

The solution is to move dynamic signaling into the pseudowire layer to give pseudowires a dynamic provisioning capability using OSPF-TE (open shortest path first – traffic engineering), IS-IS-TE (intermediate system to intermediate system traffic engineering), RSVP-TE (resource reservation protocol – traffic engineering), and LDP (label distribution protocol) to configure pseudowires efficiently over the MPLS core. Fortunately, the MPLS control plane allows this, and the PWE3 working group has defined extensions to LDP to allow it to set up and tear down pseudowires – and the MPLS control plane already supports protocols such as OSPF-TE.

From the access device it is a question of using the MPLS-based control plane protocols to first set up the tunnel, and then use LDP to signal to the other end of the tunnel – or the device at the other end of that tunnel – to set up and tear down pseudowires of different types. These are lighter protocols that are fairly simple to manage. Overall, this approach tends to make the network look like some of the older Layer 2 networks, such as Frame Relay and ATM, that have vendor-specific provisioning systems that provide dynamic setup and teardown.

Despite using the MPLS control plane, these pseudowire access devices are not Layer 3 routers. Although the use of the IP/MPLS control plane enables dynamic setup and teardown of connections, the forwarding of packets at the customer edge is still based on Layer 2 mechanisms.

Pseudowires support a large number of applications. Here’s a selection of some of the more important ones, starting with basic service types.

Ethernet over IP/MPLS
This, as shown in Figure 3, is the granddaddy of them all. The genesis of the concept of the pseudowire was the ability to take Ethernet traffic and tunnel it end-to-end across some kind of long-haul packet-backbone transport. This was driven by the idea of the extended LAN service that would allow customers to link remote locations to share in the same Ethernet LAN services.

75315_3.gifMultipoint VPLS and Ethernet pseudowires
Adding the ideas of multiple customers and multiservices to Figure 3 leads to Figure 4, which is the end goal of a multipoint virtual private LAN service on top of the Ethernet pseudowire. This uses a combination of bridges and Ethernet pseudowires. Tunnel LSPs are established between provider edge (PE) devices, and customer virtual private LANs are tunneled through the MPLS network. The result is that it looks and feels like the customer’s own LAN.

75315_4.gifSo customer A can talk to any customer A location, although connections do not always follow the same path; and similarly for customer B. To customer A it looks as if his internal network is extended across the region or world, and every location looks exactly the same.

This gives the operator a very scaleable and very manageable network. It can troubleshoot customer A’s network without affecting customer B; and customer A’s traffic is secure from customer B’s traffic. Further, the operator can sell services to customer A with levels of QOS and transport speeds different from those sold to customer B. So there is great flexibility in charging and service offerings.

Sonet/SDH access for Ethernet over MPLS
Figure 5 shows an Ethernet access circuit running over MPLS pseudowires over a Sonet/SDH access network. Ethernet is wrapped into MPLS over a Sonet/SDH access ring, comes back via various Sonet/SDH crossconnects and ADMs, and reaches the IP/MPLS edge device, from which it is switched through the MPLS core, and then exits via a reverse procedure.

75315_5a.gif“The key point is that, at the MPLS layer and above (especially for pseudowires and above), the connection runs end-to-end,” says Keith Mumford, Executive Director, Product Management and Marketing, Mangrove Systems. “It is provisioned and maintained using the same procedures and mechanisms that are available in the core network; by using standard MPLS connection routing and bandwidth reservation techniques. Similarly, the operator can diagnose faults end-to-end using standard MPLS and pseudowire-layer OA&M services such as LSP-Ping and LDP event messages transparently to the underlying network that the client service is running across.”In the same way, Gigabit Ethernet could be used as a metro access mechanism instead of Sonet. In this case, instead of the OC3(/12)/POS/MPLS/PWE3 protocol stack between the PW Access and the first IP/MPLS edge device, there would be the simpler stack of Gigabit Ethernet/MPLS/PWE3.

Pseudowires thus provide a new mechanism for using existing transport networks to make the access network packet aware, and also to use the same technology, architecture, and provisioning platform to provide these services over new, more metro-Ethernet-centric, access networks.

Pseudowires as an access solution
Figure 6 shows that pseudowires can be built on a number of different access transport media, including optical Ethernet on fiber or even over hybrid fiber coax (HFC). This creates a unified packet access network for IP and legacy service networks, including Frame Relay and ATM, and PSTN for TDM voice with Class features. It also allows backhaul of both voice and DSLAM traffic, as well as service aggregation for IP and other data traffic.

This arrangement can provide optimized aggregation of IP without the complexity of additional routing hops and also secure subscriber separation based on Layer 2. It transports multiple services over a single pipe, with dynamic bandwidth allocation between services, and all with one-touch, end-to-end service provisioning.

75315_6a.gif“The benefit is a single pipe for transporting all the services, with one-touch provisioning and no change to the customer premises, either in the equipment or in the way in which it is configured” says Steve Byars, Vice President, Marketing, Axerra Networks. “So we are really talking about extending all the benefits of pseudowires and a packet infrastructure to the access network.”

Multiservice aggregation over a metro network
Existing implementations of pseudowires tend to be on larger IP/MPLS edge platforms, so to move pseudowire access out towards the customer as far as possible will necessitate smaller devices.

“What is important here is to aggregate the traffic coming through the access network so that it connects with the edge of the IP/MPLS network over a Gigabit Ethernet handoff, rather than as a highly channelized Sonet interface that is expensive to deal with at Layer 3,” says Mangrove’s Mumford. “So we take advantage of the access network to aggregate traffic at the packet layer first, and then hand off over a POS or GigE interface.”

Figure 7 shows this approach. Various provider edge boxes sit at the access points of the network as virtual POPs, or on the customer premises themselves. Wrapped-up traffic – say, ATM, Frame Relay, or TDM – is brought back on pseudowires through the Sonet network. The basic Sonet network itself can provide TDM aggregation, and newer Sonet capabilities such as Ethernet over Generic Framing Procedure and Virtual Concatenation can aggregate non-TDM traffic at the Ethernet layer. So the Sonet network can aggregate all the mixed traffic onto a Gigabit Ethernet interface and hand it off to the edge router on a large pipe with a very low cost per bit compared to some of the existing channelized TDM interfaces.

75315_7a.gifCircuit emulation
In a recent Heavy Reading survey, 67 percent of service providers polled said they would consider pseudowires for circuit emulation, underlining the potential importance of this application. Circuit emulation is already being implemented by some carriers – IXCs, in particular, with lots of voice traffic, are keen to present a single TDM interface.

There are two levels of circuit emulation:

  • DS-3 and below: Targeted at multiservice access applications

  • Sonet OC-n: Targeted at transport applications

Standards for circuit emulation are maturing and converging, and include:

  • Requirements for edge-to-edge emulation of TDM circuits over packet-switching networks (PSNs)

  • Sonet/SDH circuit emulation over packet (CEP)

  • Structure-aware TDM circuit emulation service over packet-switched network (CESoPSN)

  • Structure-agnostic TDM over packet (SATOP)

Pseudowires are also appearing in specific types of operator networks. Here are three common ones.

MSO commercial services
“I think this application is going to be something of a surprise to people, because we are talking about enabling the cable multiservice operator to address an entirely new market,” says Axerra’s Byars. “MSO commercial services are addressed to business customers, and MSOs have used fiber to bring commercial services to some of the larger enterprise locations. But what about the small and medium businesses that they already pass with their coax infrastructures? Pseudowire is the one technology that allows these cable operators to seamlessly offer the very same services to the small and medium businesses that they currently offer to the larger enterprise locations.”

The lower part of Figure 8 shows a pseudowire access device that would be a new type of device: small, low cost, located at the customer premises and offering T1 for voice and legacy data applications, as well as Ethernet, and feeding into a cable modem. These services are brought across the HFC networks and terminate on cable modem termination systems (CMTS), where they are handed off into the IP core.

Additionally, Figure 8 shows the MSO offering the same services over WiMax (many CMTS vendors also OEM a WiMax solution) and over carrier Ethernet (most MSOs are already deploying fiber to larger buildings), as well as over HFC to small and medium-sized businesses.

75315_8a.gifPseudowires have undergone considerable testing for this application in the lab and in the field, and are proven to work with implementations of both DOCSIS 1.1 and DOCSIS 2.0. They can offer full-featured TDM voice (leveraging Class 5 services), as well as Frame Relay and Ethernet VPNs.

Multiservice backbone
Combining voice and data circuits over an IP/MPLS backbone to reduce costs, as shown in Figure 9, is a pseudowire application already being deployed. Typically, if an operator runs an OC3 voice link between two cities, and wants to add data, it has used a separate OC3 link for the data, thus doubling its transmission costs. Using pseudowires to transport circuit-emulation services allows the two separate transport circuits to be combined into one.

75315_9.gif“An OC3 circuit could cost tens of thousands of dollars per month,” says Overture’s Redden. “If I combine that into a single circuit, I pay back the cost of the pseudowire access device in a couple of days. So I get an immediate return in doing this right now.”

Cellular backhaul
The use of pseudowires in Figure 10 addresses a significant problem for mobile and cellular wireless operators. The problem is most acute in the radio access network, which backhauls the voice and data traffic from the base transceiver stations (BTSs) to the mobile switching centers, and relates to the high cost of bandwidth. U.S. wireless operators alone are spending more than $1 billion per year for this backhaul, and this will grow considerably as the industry transits to the 3G services environment, which requires a lot more bandwidth to transport the new data traffic that 3G will generate.

75315_10a.gifThe solution is to use pseudowires to enable the use of new lower-cost packet access infrastructures, such as metro Ethernet and broadband wireless radio (e.g., prestandard WiMax), and to converge both voice and data over them. Pseudowires are now proven in GSM and CDMA applications, and can deliver high-precision clocking to 3G Base Transceiver Station (BTS/NodeB) specification across the packet transport. The delay requirements and the need to extend very precise clocking or synchronization out to the cell sites are critical for 3G networks, where clocking precision is specified in parts per billion.

A strong attraction is that the pseudowire solution is easily added to the existing wireless infrastructure, interoperating with the BTS and equipment already in place, and supporting both packet- and circuit-based voice, which is a very demanding application, while providing a flexible interface that smoothes the transition from Frame Relay to ATM to IP and Ethernet for the data services. The provision of both circuit emulation (TDM) and service emulation (IP, Frame Relay, and ATM) is critical in the rollout of 3G (and higher) mobile services.

FTTH backhaul
Ethernet passive optical network (EPON) looks to be one of the key FTTH (fiber to the home) technologies, and growing amounts of FTTH are being deployed in Asia and the U.S. Figure 11 shows how pseudowires can provide converged packet and TDM access to metro hub sites around a CWDM access network, and where an integrated control plane enables multiservice pseudowire and TDM provisioning.

75315_11.gifThe key point is that, although there is a lot of Ethernet traffic running over these networks, there are also other traffic types – such as Frame Relay and TDM – entering and leaving the metro POPs. So it is a multiservice application, combining pseudowires, multiservice support, and Sonet (say) over low-cost CWDM. Being able to provide low-cost high-bandwidth access for multiplexed pseudowires over the Sonet access infrastructure is thus of very great interest for backhauling the large amounts of data that cross the EPON network. This data is then handed off to the MPLS core network, but some of the TDM services can be handed off to either the voice network or the optical network for straightforward private-line type of provisioning.

Providing the ability to set up and tear down circuits on demand, and to converge packet with TDM access at the access point, points to new types of architectures to be deployed at much lower cost in the backhaul network. This is not just in terms of the capital expenditure for the equipment, but also in terms of the recurring operating cost for leasing backhaul fiber or maintaining the equipment over it.

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like