Network Computingwww.networkcomputing.com

Network architects trying to prepare for the 21st century understand the importance of choosing the right backbone technology upon which to build a corporate network. Some sources may tell you that ATM is "dying," but premature predictions and speculation in the market will get you only so far in designing real networks.

If you look beyond the sensationalism and media hype, you will find hard evidence that ATM has a home; that home is moving away from the desktop, through the LAN backbone, and toward the wide-area carrier backbone.

Desktop Self-Wash
ATM to the desktop has been a controversial topic for several years. The bottom line is simple--no matter how easy it is to install, and no matter how many features it provides, few customers will ever go out into the workforce and replace hundreds or thousands of functional Ethernet NICs to provide ATM services. With PC vendors shipping built-in or bundled Ethernet NICs, the marketing options are grim. Although many of us believe integrated voice, data and video to the desktop would be the greatest thing since sliced bread, no one wants to tear out millions of dollars' worth of installed PC and telephone infrastructure just to make it happen. Furthermore, desktop videoconferencing doesn't seem nearly as attractive when you think about how you look in the morning before your first cup of coffee.

The business case for integrated voice and data services is straightforward: Combining voice and data networks would reduce administ rative overhead associated with running t wo separate departments for phone and PC maintenance. Unfortunately, the complexity of ATM technology--getting your emulated LANs and phone circuits to work together--requires a doctorate in ATM networking. The ideal desktop ATM connection would have been a small ATM transceiver with Ethernet and telephone connections built in. The system would have been transparent to the user, wouldn't require opening the desktop PC, and would achieve the unified network goal. But like so many other potentially useful products, this one doesn't exist. The sun is setting on ATM to the desktop, and the battle was won long ago by the Ethernet vendors. For most of us, the only way ATM will get to the desktop will be via trade magazines.

Deluxe Hot Wax
ATM in the LAN backbone is where it's at. Gigabit Ethernet is nipping at the heels of this market segment, but ATM has a big running start, so putting all your eggs in the Gigabit Ethernet basket would be premature. The $500-to-$1,500 per-port cost of ATM equipment yields a wealth of features, but evaluate these features carefully, since some add value to your network while others simply add to the price. Gigabit Ethernet will further complicate the picture with ATM-like features at a fraction of the cost. Managers looking to make a decision within the year should weigh ATM carefully, because it has the upper hand on several of the following important features:

· Scalable Bandwidth One inarguable fact about ATM technology is how well the fabric scales. This applies both to the switch backplane and to the network architecture. For example, Texas Instruments and Stanford University are teaming to build a 320-Gbps switch that could be used as the core for an ATM switch (read about it at tiny-tera.Stanford.

EDU/tiny-tera). Scalable bandwidth is not an issue. Tests from firms like Strategic Networks Consulting Inc. (SNCI) have conclusively shown that even at the fast ATM speeds, switches are capable of forwarding cells at wire spe ed. And thanks to the connection-oriented nature of ATM, scaling your enterprise backbone is as simple as plugging in another ATM link between switches. Network Computing demonstrated that feature in its April 1 review of ATM edge switches (see "Scale Ahead of the Traffic With ATM Edge Switches," at www.NetworkComputing.com/806/806r1.html). Fast Ethernet and Gigabit Ethernet vendors are boasting similar trunking technologies, but these are proprietary and interoperability is a long way off. If networks continue to grow at the pace we've seen in the past five years, scalable bandwidth will become a necessity, and an interoperable standard will become more appealing. To achieve this degree of interoperable scalability, you have to be willing to pay the price--a more-complex network based on a standardized signaling protocol, as well as its associated dollar cost.