Just about every discussion with telecom equipment vendors over the past four years seemed to include a casually dropped, “You know, that whole If-You-Build-It-They-Will-Come thing is over.” Everyone agreed that the era of big network buildouts in advance of customer demand was behind us, gone the way of Shoeless Joe Jackson and Kevin Costner’s career.

No carrier would dare lay out plans for a coast-to-coast optical backbone, a 30-metro data network, or anything sounding at all “pan-European.” Success in the carrier market would come incrementally. Every deployment was success-based; if customers wanted service, you brought the equipment to their building, plugged them in, and sent a bill. If they wanted something more sophisticated, you got your enterprise data group together and built a custom solution. What you didn’t do, and couldn’t afford to do, was roll out a whole new network based on a network planner’s recommendation that “this is going to be big.”

But the incremental, success-based model doesn’t hold up for long. If the goal is “rationalizing” finances, shedding debt, and staying the course while employees are axed and business lines sold off, then this strategy has its place. But time and the realities of a ruthlessly competitive market catch up with this approach. As the telecom sector heads into recovery, If-You-Build-It-They-Will-Come is re-emerging – this time, as the well-reasoned approach of the most conservative incumbent telcos around.

Why? They heard a message, not from a disembodied voice in an Iowa cornfield, but from their own network planners: The only competitive network is a converged network, end to end.

Translation: With three essential ingredients – IP, Ethernet, and MPLS – you can build a single network capable of carrying any service, period. With that converged network, you can compete with anyone – upstarts, invaders, or cable MSOs.

We’re already seeing the evidence of this new-found aggressiveness, and it has billions of dollars attached to it and the future of many vendors at stake. BT’s 21st Century Network (wisely named, giving BT another 96 years to complete it), AT&T’s Concept of One, and SBC’s Lightspeed are grand visions of a converged network that will require billions of dollars to complete. (See BT Faces 21st Century Dilemma, AT&T Edge Deployment Slips, and SBC Sheds Light on 'Lightspeed'.) Beyond these monster announcements, which have been met with some healthy skepticism, there are other, quieter, moves being taken, based on the same impulse.

Regardless of the size or ambition of these moves, they all have one thing in common: Ethernet is a crucial part of every plan. Whether it’s broadband access, new WAN infrastructure, VOIP, or wireless, Ethernet has emerged as the de facto data link layer for future services. The physical layer may be DSL-charged copper, Sonet, SDH, or WDM, but when it comes to moving data around, everyone has agreed on Ethernet, in much the same way as everyone has agreed on IP.

Look through recent RFPs from the RBOCs on broadband access, and regardless of whether they describe fiber to the node, premises, or curb, they are all specifying more Ethernet in the network. SBC and BellSouth have both opted for a fully converged IP access network and have made it clear that Ethernet will carry a triple-play of services from new IP DSLAMs to aggregation switches and routers. ATM, the old hope of a converged network, is hard to find in these new blueprints, having shown its limitations in the first round of broadband proliferation.

In the new era of broadband, Ethernet will eventually run from CPE to the core in ways ATM could not match. Ethernet, on its own, faces many of the same limitations that beset ATM, so MPLS is often integrated to add the missing “carrier-class” pieces (end-to-end OAM, high-speed protection, distinct service classes, network scaleability), fulfilling the goals of convergence as this technology matures.

Similar moves are underway by cable MSOs, which have spent heavily on their access networks over the past decade and are now focusing considerable energies on adding voice and improved data services for residential and small and medium businesses. The traditional digital video interface, DVB-ASI, is already giving way to Gigabit Ethernet; and packet voice and data over Ethernet are a given. Ciena’s acquisition of Internet Photonics was a clear acknowledgment of this trend: Addressing the MSO market means offering up Ethernet gear.

Where once carriers were content to plug an Ethernet blade into an MSPP to satisfy a request for an Ethernet private line, they now are being pushed to offer a full suite of Ethernet services, particularly Ethernet LAN services that extend beyond the metro or even the region to compete directly with legacy WAN services. The only way to do this is to make much bolder deployment moves, ahead of demand, with true carrier-class Ethernet infrastructure that extends networkwide.

France Telecom earlier this year announced a nationwide rollout of Ethernet infrastructure in its home country. This infrastructure will carry both packet services and emulated circuits for business customers. Atrica appears to be the winner there. Recent discussions with Verizon suggest that the RBOC is ready to adopt a similar strategy, pushing Ethernet services throughout its operating region and even out of region to satisfy customer demand for Ethernet-based MAN and WAN services. Similar strategies are now being implemented at Global Crossing and at many other operators that can no longer afford to offer Ethernet incrementally but are prepared to make the transition to a full-scale Ethernet services infrastructure.

The goal is one of migration: Get customers off Frame Relay and private lines as painlessly (and profitably) as possible while adding new customers attracted to the economics of Ethernet. Operators differ in their strategies here, some opting to let customers decide when they want to move from legacy services, while others are aggressively marketing to existing customers as a way to scale into higher bandwidths. What they have in common is a recognition that if Ethernet is meant to supplant Frame Relay and private lines, it needs to be deployed fully throughout the network. This is quite a challenge, as it requires gear that provides interworking among ATM, Frame Relay, MPLS, and Ethernet while supporting pure Ethernet services that match these services in SLAs and reliability.

The telecom bubble burst for a reason – too much competition, too little differentiation, not enough demand – so it is reasonable to ask if embracing the “build it and they will come” philosophy once again will occasion an ugly repeat of history. The risks are clear:

At the end of the day, these telcos really have no choice. The migration of services to IP, MPLS, and Ethernet has the whiff of destiny about it, one only a fool could ignore. The time of hunkering down, incremental spending, and limited rollouts is over. With all this focus on access and metro networks, it’ll only be a matter of time before we start hearing about optical network backbones again.

— Scott Clavenna, Chief Analyst, Heavy Reading

Light Reading's Carrier-Class Ethernet in China roadshow will provide an invited audience of senior decision makers from service providers in China with a unique education in how to design and deploy profitable Ethernet services, employing original research written and presented by Heavy Reading analysts.

In Shanghai on Nov. 30 and Beijing on Dec. 2, those interested may register for it here.