Can Ethernet span LAN, MAN and WAN?

Remember the initial hype surrounding ATM? Its proponents dreamed of a world dominated by ATM, the only technology able to stretch across both the LAN and the WAN, creating a single, unifying network delivering voice, video and data over the same wire to desktops everywhere.

Now, rightly or wrongly, some people are dreaming the same dream about Ethernet.

Ethernet, of course, dominates the LAN. It has scored a knockout win over token ring. And it has mounted a successful counterattack against ATM's advances. First, ATM to the desktop lost favor because Ethernet and Fast Ethernet were already well-entrenched in desktop connections. Then, ATM started losing ground in campus backbones, as Gigabit Ethernet and Layer 3 switching came to the fore. Now, Gigabit Ethernet is being used in some metropolitan-area networks (MAN), and standards bodies are readying 10-Gigabit Ethernet for metropolitan and even WANs.

Can Ethernet really pull it off? Can Ethernet make inroads in ATM's WAN turf? Can any technology really be all things to all people?

To be sure, Ethernet has some serious competitors in both MANs and WANs. ATM, for one, remains a potent force. Then there's a technology proposed by several Swedish companies, called Dynamic synchronous Transfer Mode (DTM). A third competitor is a technology developed by Cisco, called Dynamic Packet Transport (DPT).

But the most well-placed competitor is Packet over SONET, says Joe Skorupa, director of switching and routing at Ryan Hankin Kent, a market research firm in South San Francisco. "There's an awful lot of SONET out there, and it's not going away," he says.

What Packet over SONET has going for it are built-in management capabilities, which are essential for transporting data over very long distances. If something breaks down along the way, there has to be some way of letting the network's managers know. Ethernet never had to worry about those distances, so it doesn't have those capabilities.

"Over long-haul links, I still have concerns about Ethernet's robustness," Skorupa says.

One possible solution is to create a kind of "digital wrapper" around Ethernet frames for long-haul traffic, a wrapper that would have management data akin to SONET. But adding the wrapper might add complexity and expense, which defeats the purpose of using Ethernet technology in the first place.

Much of Ethernet's success has been based on its low cost and simplicity. Equipment vendors are saying Ethernet in the wide area may have as little as one-fifth the cost of SONET and one-tenth the cost of ATM.

Such numbers are attracting carriers, such as MCI WorldCom (NW, Aug. 16, 1999, page 1). The company is looking into Ethernet over long distances, but has reservations about the technology's performance in the WAN, says Rama Nune, senior manager for optical and data networks at MCI WorldCom.

10-Gigabit Ethernet

But there is enough interest that vendors have been trying to make the next Ethernet standard in the works, 10-Gigabit Ethernet, ready for the WAN.

A debate raged last year in the high-speed study group of the Institute of Electrical and Electronics Engineers (IEEE), over which speed to choose for 10-Gigabit Ethernet. The reason was that 10G bit/ sec makes sense in the LAN, but 9.584640G bit/sec is the payload rate of OC-192 SONET used in the WAN.

The group compromised in September. It decided to specify the media access control (MAC) layer of the new standard at 10G bit/sec, but it will also allow vendors to add logic to the physical layer to convert the traffic to SONET rates. That is, the physical layer will clock traffic into the LAN at 10G bit/sec, but it will clock it out into the WAN at 9.584640G bit/sec.

Because the logic would be fairly simple, adding it shouldn't have much impact on the cost of Ethernet in the WAN, says Bob Grow, an engineering fellow at Intel involved with the standards work.

The next step for the emerging standard is to get the project authorization request approved in the IEEE, and Grow says that's on track for later this month. After that, the first meeting of the official 10-Gigabit Ethernet task force will happen in March. The work is expected to receive the designation IEEE 802.3ae.

Much to be done

Putting Ethernet into the WAN isn't trivial, and the task force will have its work cut out for it, Grow says. IEEE members like Grow, who have come from a LAN background, will have a lot to learn about how to transmit data over very long distances, he says. Likewise, members who have come from a WAN background will have to learn the intricacies of the LAN technology.

Right now, the idea is that the technology would work over several hops. Here's a typical scenario. LAN traffic would be aggregated through a campus backbone switch, putting 10G bit/sec onto a long-

distance fiber-optic line. The line should be able to reach about 40 kilometers, in the same range as long-haul Gigabit Ethernet today in MANs.

The fiber-optic line would go to a telco central office, where there would be a telco repeater. The repeater is used for long-distance data hauling and shouldn't be confused with a LAN-based "repeater" in the Ethernet world, also known as a hub. The telco repeater would either map the signal onto a SONET frame or simply transmit the data on a wavelength in a dense wave-division multiplexing environment, Grow says.

Gigabit Ethernet is already a player in MANs, but whether it can parlay success in that area into success in the WAN is still to be determined. Industry watchers are now saying Ethernet technology will be one option for carriers in the WAN, but not the only one. Plus, it's not likely to be the dominant one.

But with Ethernet, you never can be sure. o