10G Ethernet WANs?

By Jeff Caruso
Network World, 08/16/99

When you think of WAN technologies, Ethernet doesn't usually spring to mind - but it may soon.

MCI WorldCom and other service providers are eyeing 10-Gigabit Ethernet as an alternative to packet-over-SONET and more traditional WAN technologies. And makers of carrier gear are pushing for specifications in the evolving 10-Gigabit Ethernet standard that would accelerate its use beyond the LAN.

For example, the high-speed study group of the IEEE is debating whether to standardize on 10G bit/sec or 9.584640G bit/sec. The latter would match the OC-192 rate of SONET, a widely used carrier network technology, making it easier for service providers to adopt 10-Gigabit Ethernet.

Carriers are attracted to 10-Gigabit Ethernet for more than LAN transport services. Ethernet has proven itself as a simple and inexpensive way to build networks of various speeds. In fact, some carriers are already using Gigabit Ethernet in metropolitan-area networks (MAN).

As a result, Ethernet could be on the verge of becoming what ATM was supposed to be - a single technology that can stretch from the desktop to the campus backbone to the MAN and beyond. Possible enterprise benefits are less-expensive services and easier network management because the network could all be based on a single core technology.

"We're looking at the impact of native [Ethernet] interfaces and how we can efficiently transport Ethernet across our network," says Rama Nune, senior manager for optical and data nets at MCI WorldCom. The company is the only major service provider to openly discuss its research in this area, though industry sources say others are looking into it.

Ethernet still falls short of ATM and SONET in many respects. Critics point out that Ethernet doesn't have ATM's quality-of-service (QoS) guarantees, which are necessary for real-time traffic. Plus, Ethernet doesn't have the built-in management capabilities of SONET to help service providers track link problems.

But equipment vendors and service providers are looking at ways to overcome the shortcomings. Ethernet can get a degree of QoS through priority queuing, for instance.

And one possibility for beefing up WAN management is to create a "digital wrapper" around Ethernet frames for long-haul traffic, says Joe Skorupa, director of switching and routing at Ryan Hankin Kent, a market research firm in South San Francisco. "When you're going across the country, you've got a complex cable plant in the middle and you need some visibility into the operation of the network," he says. "Plain vanilla Ethernet framing doesn't give that to you."

"The key question is whether we can offer maintainable services," MCIWorldCom's Nune says. Error statistics, failure conditions and other information are vital to figuring out what service-level agreements the provider can offer and how they can be enforced, he says.

The main reason Ethernet is being considered for the WAN is the low cost. Ethernet gear costs one-fifth the price of its SONET counterparts, says Dan Dove, principal engineer of LAN physical-layer technologies at Hewlett-Packard.

So why wasn't Ethernet considered for WAN transport before now? "ATM was much more viable when Gigabit Ethernet was being developed," says Bob Grow, an engineering fellow at Intel who has participated in the standards process for Gigabit and 10-Gigabit Ethernet. The conventional wisdom several years ago was that ATM would form the backbone for most corporate networks, creating links to ATM beyond the campus, he says. However, Gigabit Ethernet is becoming the preferred choice for many corporate backbones.

Technically, the standards work on 10-Gigabit Ethernet hasn't really begun. The IEEE study group is still trying to decide on objectives, such as what speed to endorse. This concerns Michael Bennett, a network engineer at Lawrence Berkeley National Laboratory, who would like to get his hands on 10-Gigabit Ethernet gear and doesn't care much about carrier use of the technology.

"Until that [speed] issue gets resolved, I'm afraid it's going to delay the whole process," he says. The IEEE study group hopes to submit a project authorization request for 10-Gigabit Ethernet by year-end.

Reaction: Here's what some Fusion users are saying about this article: What do you think? Add your comments to the thread