The Networkologist

Is Gigabit Ethernet The Next Miracle Cure?

Is Gigabit Ethernet truly the next great thing, or is its marketing designed to stall ATM so that the power brokers of infrastructure technology (namely Cisco) will hold sway into the next decade? It's too early to tell which high-speed technology will dominate, but both Gigabit Ethernet and ATM need maturation. Gigabit Ethernet standards are expected to be complete in early 1998, with prestandard, interoperable products showing up late next spring.

Network Rogaine Gigabit Ethernet is marketed like a miracle cure. According to its proponents, it melds the best characteristics of Ethernet (frame format, use of existing network architecture and millions of experienced network managers) with the high points of Fibre Channel (nearly 1-Gbps data rate and available physical lay er interfaces), and does so at two to three times the cost of 100BASE-T.

The Gigabit Ethernet Alliance says its technology is most useful for links between Fast Ethernet switches, as an upgrade to Fast Ethernet switches and FDDI switches, and on a much smaller scope as a desktop technology for people who have maxed out their desktop 100-Mbps switches. Gigabit Ethernet initially will run over single-mode fiber and multimode fiber, and the IEEE 802.3z task force is working out an answer for the millions of network managers with Category 5 installed.

Always Read the Fine Print Ask questions as you lay out your next generation of infrastructure strategy. In the spirit of market hype, the Gigabit Ethernet Alliance's Question and Answer FAQ says, "In principle, there is no limit to the size of a Gigabit Ethernet network, either in terms of physical extent or number of nodes." We can't tell principle from practical until the standards are formed and the products are shipped, but unless you're willing to work with fiber or have installed it recently, you'll have to compromise distance for speed.

With multimode fiber, you'll have a Gigabit Ethernet network diameter of 200 meters. Single mode will get you 2 kilometers. A twisted-pair repeater will get you a paltry 25 meters, so don't plan on Gigabit Ethernet becoming your standard desktop infrastructure using your existing Cat 5 cable plant.

For Best Results, Be Patient Gigabit Ethernet's apparent similarity to Ethernet and Fast Ethernet will be a big advantage over 622-Mbps ATM. The Gigabit Ethernet proponents are co-opting network managers' familiarity with Ethernet and their reluctance to adopt ATM for a variety of reasons. Gigabit Ethernet repeats the ATM chant--"scalable bandwidth and quality of service"--but adds "at a reasonable price and a familiar technology, too."

While ATM answered the call to carry voice, video and data over a single transport, it comes to the door virtually empty-handed of applications that take advantage of A TM's quality of service features. Winsock and native ATM APIs are coming. Meanwhile, the Ethernet community says it can do everything ATM can: Couple switched Ethernet (in its 100-Mbps and 1-Gbps incarnations) with 802.1q to explicitly determine application priority for virtual LANs, RSVP in the routers and RTP in the applications, and an Ethernet-based network can do everything that ATM has promised for so long--but at a better price. Gigabit Ethernet plays to an old strategy in networking: over-engineer the network. Give your users and backbones as much bandwidth as you can afford because in two years, their requirements will outstrip your wildest estimates.

New technologies promise more than they can deliver, but what Gigabit Ethernet promises looks good. Whether it or 622-Mbps ATM gets implemented as the switch interconnect or on the backbone depends on both technologies' standards being completed, interoperable products that are priced aggressively and network managers' desire for quality of service balanced against applications' need.

Patricia Schnaidt can be reached at pschnaidt@nwc.com.