Bringing ATM closer to reality

Communications News, Sept, 1994 by Alan Stewart

Asynchronous Transfer Mode (ATM) is a fast packet technology designed to combine voice, data and video in fixed length cells and transport them from the user premises across the network using standard interface and switching protocols.

Several users have tested ATM. The Johns Hopkins School of Medicine is running a demonstration with LightStream Corp., Billerica, Mass., and UB Networks, Santa Clara, Calif., checking ATM's applicability for advanced medical imaging.

LightStream 2010 switches connect Hopkins and the University of Maryland Medical Center across town via a leased T1 line. UB Access/One hubs connect several Silicon Graphics archival workstations to the switch at Hopkins via an Ethernet connection to routers. Scanners at UM's radiology department are connected via Ethernet to a router that feeds the 2010 switch at that end. But this is not an everyday application.

ATM's principal problem at present is that it's too fast and too expensive. Most industry observers agree that demand for ATM will occur only when it can be used to bring real cost benefits to everyday local area networking operations.

Ralph Ungermann, founder and former CEO of Ungermann-Bass, may have brought this possibility much closer. He announced on May 31 that his new company, First Virtual Corp., Santa Clara, Calif., has signed a licensing agreement with Advanced Telecommunications Modules Ltd. (ATML), of Cambridge, England. ATML, a spin-off of Olivetti Corp., has been developing low-cost ATM that supports desktop multimedia networking.

Using the new technology, which apparently pumps data at a much lower rate than existing ATM products, desktop connection costs could fall dramatically. This holds the promise of a dynamic marketplace for applications that depend on switching multimedia such as desktop video teleconferencing. FVC plans to announce product details at a major computer show in September.

"FVC will be delivering ATM networks for less than $500 a connection that are compatible with the existing LAN environment and twisted pair office wiring," says Ungermann. This cost per connection includes the adapter card in the PC or workstation and the port or switch in a wiring closet on user premises (equivalent to 10Base-T).

"As more and more integration occurs, costs drop automatically, and our technology can accelerate this trend, creating a strong marketplace," says Jeff Leffingwell, co-founder of FVC. "Existing ATM solutions run at a higher speed than Ethernet, for instance, so the cost of wiring and peripheral equipment may be higher." Pumping ATM at slower speed brings down costs and still provides the user with a high-speed multimedia capability.

The ATM picture is further clouded by the fact that a full set of standards for ATM is still about two years off. This inhibits vendors from maximizing the benefits of incorporating full integration techniques. The other problem is that users have existing installed bases of LANS and peripheral equipment, Leffingwell explains.

Lower speed ATM can emulate existing LAN at a throughput of 25 Mb/s and work over Category 3 wiring in the horizontal level. The benefits are immediately obvious and include full duplex at 25 Mb/s and no collision problems. Comparing this with actual LAN throughput makes this picture even better--in the real world this is about 40% of rated 10 Mb/s speed.

The other driver is multimedia applications. Video teleconferencing, for instance, is difficult to do on Ethernet and is limited to two people. Multimedia data communication disrupts other data activity leading to conventional LAN users being frozen out. Because ATM is scalable, four or more people can participate in the conference. Because existing LANs will be remain in service for many years, it is necessary for ATM to work in this environment.

"This is why we consider cost to be the key factor," maintains Leffingwell. "The Yankee Group has published data (see graph: LAN Technology Forecast) that includes cost-per-seat-per-megabit, cost of adaptor and cost of switch port. To see the true benefit of ATM, these costs must be divided by the increased data throughput. We will announce product details in the middle of September and launch at a computer trade show."

[CHART OMITTED]

CIMI Corp., Voorhees, N.J., has just completed a study on user demand based on specific price points, consultant Tom Nolles says. "Our results made it obvious that ATM end-user demand is totally a price issue. Users are prepared to pay 25% more for ATM provided they can keep it longer and thus depreciate the cost of the networks over a longer period. As far as speed is concerned, 155 Mb/s ATM connection is a throwaway because no one needs that kind of speed. A much more realistic speed is 20 Mb/s."

The FVC projected price of $1,200 ATM for a 25 Mb/s connection by next year falling to $500 by 1996 is intended to forestall IBM, which has presented the ATM Forum with its product plans for ATM to the desktop. "With its resources, IBM is much more likely to bring product to market over the next 18 months," Nolles adds.