"Never bet against Ethernet" – Ethernet's ability to prevail over seemingly superior technologies has made this into a networking axiom. Ethernet won in the LAN hands down; it won in the WAN, spelling a slow doom for the dominance of Frame Relay and ATM; and in the MAN it created an entire niche for itself as a high-speed business service for Internet access, VPN access, and transparent LAN connectivity.

Next in Ethernet's sights: the transport network.

While most carrier Ethernet services today continue to run over a Sonet/SDH transport infrastructure, more and more operators are beginning to look at running Ethernet natively over an Ethernet PHY, first in specific areas of the network, but ultimately end to end.

The great Ethernet transition is proceeding along multiple vectors at once. The first is carriers' continuing transition from DS3 and OCn network interfaces to Ethernet. On routers, a move from channelized TDM interfaces to Ethernet interfaces can often result in a 75 percent reduction in equipment costs. The use of generic framing procedure (GFP) and virtual concatenation (VCAT) has allowed Sonet/SDH gear to add native Ethernet interfaces to multiservice provisioning platforms (MSPPs), as well.

Next is the transition from services that rely almost exclusively on TDM as the bearer (private lines, Frame Relay, ATM) to Ethernet-based ones. This is largely driven by the concomitant transition to IP, which fits most comfortably within Ethernet. As services such as IP VPNs, dedicated Internet access, and VLANs took hold, Ethernet has become the de facto method of access and metro transport.

Third, there follows the transition of networking from the STS or VC to the VLAN or pseudowire. In instances where voice is packetized and data services converge around IP, the unit of currency in networking becomes the packet flow rather than the circuit. These packet flows are networked as VLANs, MPLS label switched paths (LSPs), or pseudowires, providing a multiplexing hierarchy analogous to Sonet/SDH, although not quite as well-defined and rigid.

Lastly comes the transition of the physical layer from Sonet/SDH PHYs to Ethernet, G.709 OTN, and WDM. This is, in many ways, the final stage in the transition to an optical transport network in which Sonet is almost entirely absent and Ethernet provides all the networking elements. This transition is occurring today in pockets of operator networks, such as metro Ethernet networks, Ethernet access networks, or core networks employing a 10-Gigabit Ethernet LAN PHY directly over DWDM, but there is real industry momentum toward supporting an end-to-end Ethernet optical transport network (EOTN) that can dispense with Sonet/SDH.

This transition will be far from instantaneous: There are so many Sonet/SDH network elements in operation worldwide that Ethernet cannot possibly force them all into early retirement at once. Yet our latest research, published in the Heavy Reading report Sonet/SDH-to-Ethernet Migration Strategies, demonstrates that carriers are starting to get serious about moving beyond their current generations of Sonet/SDH gear to better accommodate what everyone agrees will be a torrent of new packet traffic coming from business Ethernet and residential triple-play services.

Consider these findings from our global survey of more than 130 network operators, conducted exclusively for this report:

Ethernet's increasing presence in the access network, particularly as a solution for triple play and IPTV, is forcing carriers to rethink their transport networks. Take AT&T Inc. (NYSE: T), for example.

AT&T's Project Lightspeed is based predominantly on a fiber-to-the-node architecture (except in greenfield deployments, where it will bring fiber directly to homes). In this network, all communications services to the home are packetized, including voice, and delivered over a mix of fiber and VDSL. The connection between the home and central offices is pure Ethernet, top to bottom, as Ralph Ballart of AT&T (then SBC Laboratories) made clear in a presentation at Heavy Reading's Links 2005 Executive Summit in November. Ballart emphasized that the network underpinning this rollout would be all Gigabit Ethernet, from DSLAMs to Ethernet aggregation switches to the IP edge.

Sonet will still play a large role in AT&T's overall network, but when it comes to next-generation access infrastructure, Sonet is absent. Our conversations with other major incumbents worldwide often turned up similar plans. Few operators are talking about a network-wide, simultaneous transition to Ethernet transport, but where new optical networks are being deployed (access, metro, regional, or long haul), Ethernet is most often considered a valid, if not favored, alternative to Sonet/SDH.

Is there a catch? Performance management will doubtless dog Ethernet for a few years to come. In our survey, an overwhelming majority of network operators said Ethernet transport must match the performance of Sonet/SDH to meet their network requirements. This is a tall order, considering that many Ethernet operations, administration, and maintenance (OAM) standards are still in their adolescence.

Carriers also expressed some reservations about the viability of circuit emulation across large packet-based transport networks. Circuit em is often performing quite well today in isolated areas, such as access loops, but many operators said they weren't yet quite comfortable with what they have tested.

Furthermore, one must always be wary of the disparity between an operator's wishes and the reality of technology maturity. Many network planners and operations engineers are greatly enthused about Ethernet transport, but they are finding the details nettlesome. How do carriers interconnect securely between EOTNs? What are the standard performance metrics to gather in support of service-level agreements? Where does OTN (G.709) fit in this picture? Will there be a need for "transition" MSPPs, such as the 1850 TSS recently released by Alcatel (NYSE: ALA; Paris: CGEP:PA)? Or will it make more sense to just build an EOTN overlay and slowly phase out the Sonet/SDH network?

When pinned down, most operators remain quite cautious. If an overarching strategy (such as Project Lightspeed) is in place, it becomes easier to talk about pure Ethernet – especially when you consider that this project basically involves Gigabit Ethernet connections between Alcatel gear, so standards compliance isn't quite the big issue it could be. Considering the billions of dollars of Sonet/SDH gear out there already, folks in charge of those networks tend to like the idea of new line cards or new MSPPs with better packet handling.

It may come down to which axiom you believe has the greatest heft in this market: "The death of Sonet/SDH has been greatly exaggerated," or "Never bet against Ethernet." The numbers suggest the latter, but my gut tells me Sonet/SDH will be doing just fine for another five years.

— Scott Clavenna, Chief Analyst, Heavy Reading