Gigabit Ethernet services aren't just cheaper; they're also easier to deploy while offering higher bandwidth and enhanced features. And most IT managers deal with Ethernet every day, so the technology is familiar.

Before the advent of Ethernet over fiber, however, copper wiring's distance limitations made the technology unsuitable for wide area and metro-area networks. Category 5 wiring can handle only 100 Mbps of Ethernet traffic, and it's limited to a distance of only 100 meters, making standard Ethernet poorly suited even for networks that connect adjacent buildings. With Gigabit Ethernet over fiber, higher speeds and longer distances became available for metro access, though Gigabit Ethernet can't meet core network needs, because WANs demand even more bandwidth.

Further, Ethernet has always been a "first-come, first served" technology, in which all data are treated equally. Managers can't determine who gets what. Sonet and ATM, in contrast, create circuits or channels to carry particular traffic types bound for specific destinations.

Sonet and ATM both can transmit data at OC-192 speeds, almost 10 Gbps. (Frame relay, also in the carriers' domain, is most economical for relatively low bandwidth needs of 10 Mbps or less.)

The Bells and other established carriers have been selling their ATM and Sonet services with MPLS (Multiprotocol Label Switching) technology, which lets IP packets travel over switched networks with minimal fuss. By using MPLS as a stopgap, carriers can leverage the networks they have in place while building new networks designed around Ethernet and IP services.

The main advantage of MPLS is its ability to prioritize traffic. MPLS tags each packet depending on its destination and bandwidth requirements. The underlying Sonet or ATM technology assigns the tag defining the route based on current network usage and traffic priority. So class and quality of service can be guaranteed more easily in an MPLS network than in a pure Ethernet setup, which does nothing to guarantee either.

MPLS services are delivered to individual customers as a private-line connection to the carrier's network. A router is connected at the customer premises to direct traffic. The customer router can be MPLS-aware--that is, able to tag packets--or it will merely deliver packets to the carrier network for tagging and delivery to their final destination.

The ATM and Sonet solutions are sound, but they, too, come with many caveats. Foremost, ATM is a totally different technology from Ethernet, so it presents a difficult learning curve for most IT professionals. The other drawback to ATM--and to Sonet--is the expense. ATM has gotten less expensive over time, but it hasn't seen the enormous price drops that Ethernet has. ATM and Sonet equipment capable of OC-192 speeds costs upward of hundreds of thousands of dollars, while 10 Gigabit Ethernet costs only tens of thousands. Furthermore, 10/100-Mbps Ethernet cards have come down from hundreds of dollars to around $10 each. If OC-3 NICs, which transmit data at 155 Mbps, cost less than $20, rather than, say, the $600 it costs for a typical OC-3 NIC from Marconi, the world might have a lot more fiber deployed to the desktop by now.

Although Sonet has been considered by some as an alternative because it avoids the 5-byte (10 percent) "ATM tax," its price too has also kept it out of all but the largest of corporate budgets. Furthermore, Sonet switches aren't equipped with large buffers, so the traffic needs to be groomed before it hits the switches. This adds a layer of complexity with which managers shouldn't have to deal.