Gigabit Ethernet hits the WAN

By Jim Duffy
Network World, 7/20/98

Spokane, Wash. - The benefits of Gigabit Ethernet in the LAN are well documented: high-performance, congestion-free connections for enterprise backbones and server farms - all with the familiarity of good old Ethernet.

But Gigabit Ethernet in wide- or metropolitan-area networks? Unheard of, you say. Everyone knows Ethernet isn't a WAN technology.

Don't tell that to the 13 Washington state educational institutions that this week will announce plans to build a wide-area Gigabit Ethernet network supporting voice data and video traffic. The network will use dark fiber provided by a local electrical utility and network gear from Packet Engines, Inc. The project illustrates a compelling new application for the technology: Customers could build high-speed, multimedia Ethernet WANs that cost less and provide greater bandwidth than today's inflexible T-1 or T-3 networks.

For vendors of ATM products, the Washington state deal also serves as a warning that Gigabit Ethernet companies such as Packet Engines and Extreme Networks, Inc. aren't content to remain LAN-locked. According to Bernard Daines, president of Packet Engines, wide-area Gigabit Ethernets are the result of three different forces: the availability of low-cost, wire-speed routing devices; the industry's convergence on IP; and the availability of dark fiber and alternate long-distance channels.

"That allows people to think of networking in a very different way," Daines said. "They can consider their whole operation as an extended LAN, rather than having to go to some carrier that carries bits at a lower rate in between LANs."

Packet Engines, based in Spokane, Wash., is already building the Washington state network - called E-MAN - and another network like it, and claims to have several more projects waiting in the wings. Extreme, of Cupertino, Calif., is involved with a trial in Great Britain in which telecom giant British Telecommunications plc (BT) is evaluating a multimedia service based on Gigabit Ethernet.

But don't get too excited. Even supporters said Gigabit Ethernet metropolitan-area networks and WANs are concepts in their infancy, and the E-MAN and BT projects are on the bleeding edge of technology.

Esmeralda Silva, LAN analyst at International Data Corp. in Framingham, Mass., said she isn't expecting a big boom in Gigabit Ethernet metropolitan-area networks or WANs any time soon. "ATM is still really the technology for larger accounts that are looking for something that reaches out into the wide area." The key reason: quality of service. Gigabit Ethernet cannot provide the same levels of voice and video quality that ATM can, and nobody expects that it will, Silva said.

Daines disagreed. He said Gigabit Ethernet not only provides enough capacity to overcome many quality-of-service (QoS) issues, but vendors are building new features - from queuing mechanisms, traffic prioritization schemes and management functions - into their devices that can ensure QoS. He added that the E-MAN project will prove Gigabit Ethernet can provide adequate QoS for voice and video applications. "The schools recognized this a while ago," Daines said. "Everybody else bid ATM; we bid Gigabit Ethernet and were selected for a fairly large operation. People are finding that there really isn't much advantage that ATM offers, and ATM is more expensive."

BT is also discovering that Gigabit Ethernet can deliver QoS. The carrier is using Extreme switches in the backbone and at the edge of a campus testbed network of 1,000 users. The carrier is running Cisco Systems, Inc.'s IP TV software to enable those users to participate in audio and video conferences.

"It all works like it's supposed to," said Mark Salter, network systems engineer at BT. "I've saturated a gigabit link and tried to put a video stream through it with and without quality of service. Without QoS, you don't get anything; with QoS, it works."

The major limitation to Gigabit Ethernet in the metropolitan-area network and WAN is distance. Users can run Gigabit Ethernet frames and packets up to 100 kilometers without repeaters, but it's unlikely that the technology can run around the globe natively, Daines said.

"We're not going to get 3,000 miles in one hop," he said.

How carriers and service providers perform amplification, repeating and passing of packets from point-to-point will help determine how far Gigabit Ethernet can go in the metropolitan-area network and WAN. Daines said he believes that the existing packet-over Synchronous Optical Network (SONET) infrastructure and services will go a long way in extending Gigabit Ethernet over great distances.

Another factor in determining Gigabit Ethernet's feasibility in metropolitan-area networks and WANs is access to dark fiber, a term that refers to fiber cable that is leased without switches, optical transceivers or services. Today, private fiber facilities are scarce.

Indeed, some observers say the E-MAN project is a better example of a dark fiber application than it is a proof of concept for Gigabit Ethernet in the metropolitan-area network.

"If you can get access to dark fiber at a reasonable price, and if you're within the distance limits, then the economics are wonderful," said Ron Jeffries, president of Jeffries Research in Arroyo Grande, Calif. "But does it ever make sense if there's no dark fiber? I don't think it does today."

Without dark fiber, users would have to rely on a service provider offering some kind of native Gigabit Ethernet service. That's not a big topic of discussion today, but if there's enough risk to the current T-1 and T-3 business, carriers may see more of an incentive to offer the high-speed LAN technology in metropolitan-area networks and WANs.

Until then, E-MAN pioneers will be creating the business case for both Gigabit Ethernet WANs and dark fiber.

"People can build their own networks," said Packet Engines' Daines. "They don't have to provision a network by leasing service from the carriers. All they have to do is lease fiber, which is very much like leasing a building."