How Far Can MPLS Go?

MPLS is now so well established in carrier IP network cores, there is little room left for debate over whether it will be trumped by possible alternatives. Performing as advertised, MPLS is letting operators traffic-engineer their IP backbones, reliably support network-based IP VPN services, and trunk legacy traffic such as Frame Relay, ATM, and Ethernet using MPLS-based "pseudowire" technology across underutilized IP cores.

This all adds up to improved network performance and sustainable profitability for packet-based services. And it's putting billions into the coffers of router vendors such as Cisco Systems Inc. (Nasdaq: CSCO) and Juniper Networks Inc. (Nasdaq: JNPR), which have led the MPLS charge in the edge and core.

But MPLS isn't staying put, driven by the novel thinking of a few service providers that are looking to expand the role of MPLS and pseudowires into metro transport gear and even access equipment. It sounds quite a lot like 1992, when ATM was called out of the core and offered up to the edge, access, and even the desktop (25 Mbit/s, right into your PC!). New networking technologies tend to follow a similar path – success in one area will breed (hubristic?) ambition to solve every networking challenge, end to end.

A host of hungry vendors inevitably wait in the wings, hoping this may be the opening they have been waiting for – a slot in the convergence market where they won't have to face the networking Goliaths head-on.

I recently dug into this burgeoning "pseudowire solutions" market and found quite a loud buzz out there, with vendors using pseudowires in ways far beyond what Luca Martini first envisioned.

First, some background: Operators, particularly IXCs, have spent the last few years looking at ways to extend MPLS further out into metro and access networks. AT&T Corp. (NYSE: T) and Sprint Corp. (NYSE: FON) have published papers on the idea, and MCI Inc. (Nasdaq: MCIP) has made explicit mention of pseudowire technology in the descriptions of its new "Converged Packet Access" network. The value of MPLS in metro and access networks, these operators say, is not so much about traffic engineering as it is about network consolidation (capex and opex reductions) and new services (VPLS, IP VPNs).

Convergence via MPLS in the access net means you can hand a customer a single connection (a DS-3, Fast Ethernet, OC-n, it doesn't really matter), adapt all their traffic into MPLS-based pseudowires, then carry it back to the IP network edge for further treatment. The nice thing is, it will also allow you to perform some statistical multiplexing, add a new level of MPLS-based service protection and security, and unify the management of a diverse set of services. It can also help reduce your router and switch port usage in major POPs, saving on precious capex, and address some costly areas of opex by easing service initiation. And, for IXCs, it provides another tool to reduce or bypass ILEC access charges, the perennial killer of their bottom line.

For RBOCs and PTTs, the appeal of MPLS in the metro and access network isn't quite so clear, though packetizing all services at the customer edge can have regulatory benefits to some. This trend will also, ultimately, dovetail with their own ambitions to drive outside their regulated regions and become national operators, either via acquisitions of IXCs or through new metro network builds.

Among incumbents, BT Group plc (NYSE: BT; London: BTA) seems to have the best sense of how to extend MPLS beyond the IP network edge. Trailing the British giant are a number of Asian operators, already well on their way. In the U.S., the RBOCs are taking their time, still evaluating VPLS as a next step beyond their current Transparent LAN offerings. None has made the bold assertion that all services (voice, Ethernet, Frame Relay, ATM, etc.) will be converted to pseudowires in the access network. That's clearly too disruptive a step for their taste.

Beyond the RBOCs, a variety of alternative operators, cable MSOs, and wireless operators are also seeing the value of moving to pseudowire solutions, most often for the same reasons as the IXCs – streamlining services management, reducing backhaul costs, and unifying operations.

The vendor community is gearing up for this move to end-to-end MPLS, and venture capitalists are backing several solutions, as shown by today's announcement of Mangrove Systems Inc.'s second funding round (see Mangrove Lands $21M Round). After studying the pseudowire solutions market for six months, a landscape of new product opportunities began to emerge, based on service provider requirements for migration of their Layer 2 services to an IP/MPLS infrastructure and improvement of their metro and access network efficiency and operations with pseudowires.

The heritage of current pseudowire solutions vendors includes the following:

  • Ethernet access solutions. A group of vendors offering products that adapt customer traffic to Ethernet VLANs for backhaul to a central office or POP. Migrating this product class to pseudowire/MPLS adds improved protection options, OAM capabilities, standards-based multiprotocol encapsulation and tighter integration with the core IP/MPLS network.

  • RPR metro and access gear. Many RPR-based metro optical operators have adopted MPLS as a path layer, using pseudowires as the means of service encapsulation and connectivity within the packet ring. RPR's value here has been attractive to a wide range of operators, though major incumbent LECs have yet to make sizable commitments. Currently, cable MSOs, CLECs, and a range of Asian operators are the most avid users.

  • Multiservice IP concentrators. Many of these employed proprietary or pre-standard multiservice techniques, and have now evolved to adopt the full suite of pseudowire specs. These function as standalone data devices aggregating customer traffic and adapting it to IP/MPLS-based networks, including metro Ethernet networks.

  • ATM/Frame Relay multiservice switches. This is a natural choice, as multiservice switches must now show a path to MPLS migration in order to stay relevant and participate in operator convergence strategies. Adopting pseudowire technology allows these platforms to integrate with operator IP/MPLS cores, while adding the valuable features of Layer 2 service interworking and signaling interworking, among other benefits.

  • Digital crossconnect and core Sonet/SDH switch. This product category, currently seeing very soft growth for pure TDM crossconnect applications, has been focused for a few years on adding more data-awareness and network-edge functionality. Key requirements for this segment include metro Sonet/SDH ring aggregation, Ethernet VLAN aggregation and grooming, and, in a few rare cases, ATM aggregation and switching. As pseudowire technology moves into the metro and access networks, these systems are sometimes being repurposed to serve as "multiservice crossconnects or gateways" providing pseudowire aggregation, grooming, and handoff to core optical and MPLS networks.

From this legacy I count four new product categories, still in their nascent form, but already showing some life in a few carrier RFPs.

Table 1:
Pseudowire Access Multiplexer A remote provider edge device, low cost, designed to take in all customer traffic (TDM, Layer 2) and map it into pseudowires for backhaul. RSVP-TE and LDP used for end-to-end service set-up. Also, VPLS demark functionality. Price: sub $10k list Axerra, Mangrove Systems, Overture Networks, RAD Data Communications Many startups, few large incumbent players here. Very low-cost device, mainly suited for IXCs and some wireless operators. Typically paired with hub device for head-end concentration.
Pseudowire Aggregation Switch Purpose-built device that aggregates pseudowires or attachment circuits from the access and metro networks and hands them to an edge router for service-specific processing. May perform PW switching or "stitching" and Layer 2 service interworking. Price: variable, but from $80k to $200k list Lucent, Nortel, Hammerhead Systems, Mangrove Systems, Tellabs Positioned as a next-gen multiservice switch, though distinct from legacy ATM/FR switches in its use of pseudowires and the MPLS control plane for L2 services migration to IP backbones. Nebulous category, with overlap among edge routers, multiservice switches, and Ethernet switches with MPLS support.
Packet-Aware MSPP/Packet ADM Metro Sonet/SDH gear that goes beyond simple Ethernet over Sonet transport by integrating pseudowire technology. Supports any-service-any-port (ASAP), multipoint connectivity (VPLS), and much broader range of services aggregation and transport. Price: sub $90k list Alcatel, Cisco, Luminous Networks, Corrigent Systems, Marconi, Native Networks RPR players conceptualized this category. Few incumbent players (except Alcatel) moving in this direction today.
Multiservice Edge Gateway This device is in effect a hybrid digital crossconnect/pseudowire aggregation switch. The value is both TDM bandwidth management and PW aggregation occurs in the same platform for efficient services grooming at the network edge. Price: highly dependent on configuration, switch fabric, etc. $90k to $350k+ Alcatel, Ciena, Mahi Networks, Mangrove Systems Evolution of the digital crossconnect. Still very early. Most DCS vendors are just now integrating Ethernet VLAN switching. Pseudowire support will be driven by use of pseudowires in the access network.

Of these four new products, the packet-aware MSSP and the pseudowire access multiplexer have some real traction in the market today. This makes perfect sense: They address real carrier needs (reduced access charges, improved metro services transport) and easily deliver some bang for the buck. Of the other two, the pseudowire aggregation switch has merits but is a much more complex product, often incorporating both an ATM and an MPLS control plane, new levels of pseudowire OAM, protection, and switching – all rather immature solutions for such a critical part of a carrier network. Vendors such as Hammerhead Systems Inc. are making a go at this space, hoping that Lucent Technologies Inc. (NYSE: LU), Ciena Corp. (Nasdaq: CIEN) (WaveSmith), Alcatel (NYSE: ALA; Paris: CGEP:PA), and Nortel Networks Ltd. (NYSE/Toronto: NT) underwhelm operators with their latest multiservice switches.

The multiservice edge gateway, however, may be the product with the toughest row to plow. This unit sits near the edge of the core MPLS and transport networks and claims intelligent aggregation and switching of both TDM and packet services. The idea is a bold one – build a kind of "god box" that consolidates a number of networking functions in one chassis, with the goal of reducing TDM crossconnect, router, and switch ports. It's little wonder that this idea has been on diagrams at some major network operators for years but never actually deployed. The closest carriers have come to this vision would appear to be Mahi Networks Inc.'s Mi7 – which does not have true pseudowire support, but does perform Ethernet and TDM aggregation at the network edge, setting the stage for this more advanced concept as time passes.

It's not hard to yield to the impression that MPLS and pseudowire solutions represent the only alternative to legacy networking, but already both carriers and vendors are offering up some alternatives. Ceterus Networks is offering an interesting TDM-based approach to the multiservice access question, while Nortel is arguing that MPLS should stay a bit closer to the core for Ethernet services, allowing a MAC-in-MAC approach prevail beyond the network edge.

So back to the original question: Is this ATM all over again? Well, yes, it does seem so, and the market will have to start weeding out these players and solutions as convergence forges ahead. Plenty of operators have told me they simply are not comfortable extending MPLS out to the customer. (One went so far as to say, "This is LANE all over again!" I confess I had to look that one up in my dust-covered, long-forgotten Broadband Networking text.) Fears of excessive cost, overwhelming complexity, scaleability hassles, and ignorant tech support will keep this market from ramping too quickly, but the success of MPLS will undoubtedly keep pushing at the network edges.

— Scott Clavenna, Chief Analyst, Heavy Reading

For more on this topic, check out the Heavy Reading report Pseudowires and the Future of Transport and Access Networks, written by Scott Clavenna and providing a complete survey of the developing market for pseudowire solutions.

Be the first to post a comment regarding this story.