Access charges. Local access charges that CLECs or IXCs must pay to ILECs total in the billions of dollars annually and have long been the profit eater of the IXC. In cases where carriers do not have dedicated facilities to a customer, they want to get as many services delivered over a single-leased connection as possible. In the recent past, ATM offered a nice solution: pack voice and data onto a single T1 using an integrated access device (IAD). Some CLECs and IXCs did alright with this solution, but its lack of scaleability, both economic and operational, relegated it to a diminishing niche status.

Why pick Ethernet? With Ethernet, a gigabit-speed connection can be carved up into high- and low-grade service channels with statistical multiplexing and link-level operations, administration, and management (OAM). In short, Ethernet is becoming carrier class, and that means it can become the universal connection jack through which customers connect to all their communications services.

VOIP as killer app. It's funny how VOIP keeps coming up when carriers talk about Ethernet access. The argument for Ethernet access has long been more bits for the buck. However, early attempts to become profitable without this model failed as competition drove an unsustainable price war. These days, when IXCs talk about converging access via Ethernet, it's not to sell you a cheaper alternative to a T1, it's about supporting the flood of multiservice packet traffic coming from the customer. VOIP has emerged as central to the future of IXCs (particularly AT&T). The keynotes at Light Reading Live events illustrated why Ethernet access must evolve to become the data-link layer for all their next-gen voice network designs.

AT&T's Richard Klapman, group manager of Ethernet and INCS, put it best: "What is a softswitch but an Ethernet switch with voice applications running on it?" In that light, would you want access over anything other than Ethernet?

Network simplification. This is by now familiar justification for all things Ethernet. MCI's Steve Jackson, director of network architecture and standards, made the case at LR Live that deploying separate access lines for each service was not just costly (see access charges above) but operationally messy as hell (see "before" illustration below). The concept of CPA is an operation manager's dream, it would seem. Just plug the customer into the "cloud," signal for an endpoint, and voila, you have a service. If the customer wants to add another service such as VPLS or VOIP transport, call up a user interface, identify the endpoints, click here and there, and start billing. And, if this nirvana doesn't pan out quite so easily, CPA still offers the benefit of a very high-speed interface to the customer that can accommodate service growth without the addition of new hardware.





So how do we get there from here? MCI's Jackson dove into technical details, describing the need for four network elements. Here they are:

1) The building aggregator, or a little "pizza box" that aggregates customer traffic into Ethernet, then maps that into GFP or X.86 connection depending on the uplink (aggregation is likely to be VLAN-based today, but will evolve to MPLS-based pseudowires in the future);

2) A multiprotocol crossconnect, located in the MCI point of presence, which comes with integrated TDM and packet switching, for multiservice grooming and handoff to the interoffice facilities, or metro rings;

3) An optical add-drop multiplexer for scaleable IOF transport;

4) A service network edge router, loaded with Gigabit-Ethernet interfaces, MPLS capabilities, interworking functions, and all the other packet-based tools necessary to facilitate convergence.

The goal of this architecture is to get to packet-based services as early as possible. Once packetized ("inside the cloud," so to speak), traffic has been decoupled from physical facilities. This is something that Gady Rosenfeld of Corrigent Systems Inc. has been driving home. It's a critical point to understand when discussing the values of Ethernet, MPLS, and converged access.

AT&T's Klapman says the industry needs to provide 100 percent access, and that means making Ethernet-access solutions cheap enough to deploy everywhere, not just for Fortune 500 companies. It also means that carriers need to cooperate to achieve the same level of interoperability with Ethernet that they have with their TDM networks. He also says the industry needs to solve VLAN complexity. Though Ethernet gear has come a long way towards achieving the stamp of "carrier class" there remain considerable challenges in implementing carrier-Ethernet security, multi-vendor QOS, scale, OAM service features, and economics.

Regardless, it's now clear from the presentations at our recent live events that the universal move to Ethernet is finally happening, at least in the IXC world. In the world of CPA, the operator is looking for a way to say, "I will hand you an Ethernet interface and connect you to any service you agree to pay for."

At least that's what I thought, until someone took me aside during a break and said, "They can't really afford to do this, can they?"

I grabbed a croissant and said, "We've got a conference for that, too. Come see us in December."

(Light Reading's Telecom Investment Conference is happening in New York on December 15. Stay tuned for details.)

— Scott Clavenna, Chief Analyst, Heavy Reading