ATM chip ties terminals to campus-wide Sonet rings
An asynchronous transfer mode, or ATM, chip, may be used with add/drop multiplexers to connect terminals to public networks and campus-wide synchronous optical network, or Sonet, fiber-optic rings. The chip may also support local area network speeds and data communications T1 rates at 1.554 megabits per second over carrier networks.
Shelton, CT-based Transwitch Corp., the chip`s developer, calls the chip an ATM switch-on-a-chip. It supports 1-gigabit-per-second connections to campus or metropolitan area network rings. This chip can also handle 10 local lines at the DS-3 rate of 45 Mbits/sec, which can be concentrated onto these rings over one or more access paths at OC-3 speeds of 155 Mbits/sec.
The company foresees additional fiber-optic applications for the chip, including
Stackable hubs: networking hubs to build local fiber-optic networks linking user terminals through hub-based switching cards.
Front-end concentrators: networking concentrators to combine moderate-speed ATM lines into one broadband data stream for high-capacity routing.
Cable-TV distribution: to bring video-on-demand to subscribers and help cable distributors provide data communications and Internet-access services.
Each cellbus user bus interface transceiver, or Cubit, chip includes a built-in cellbus backbone, with on-chip transceivers for 32 ports and support for hot-swapping of attached network-interface cards. This architecture provides a total ATM throughput of 1 Gbit/sec.
According to Kevin O`Neill, vice president for research and consulting at the Business Research Group based in Newton, MA, "Transwitch`s Cubit chip is the first with full ATM implementation. It includes the important functions of ATM address conversion, bus input/output and cell prioritization queuing. The latter is necessary to maintain quality of service, especially in multimedia applications."
O`Neill believes that the Cubit chip may make it easier and cheaper to enter these equipment markets. "All the players in the networking industry--manufacturers of adapter cards, hubs, routers and access providers--are factoring ATM switching into their future products.
"It`s perhaps the first time in the history of premises networking that equipment manufacturers are saying the same thing as network service providers. ATM may have a pervasive impact across the industry, and the Cubit chip could substantially reduce the cost of designing and building these systems. Transwitch is an early and strong player in the ATM-chip market," O`Neill adds.
But Transwitch is not the only player, according to Peter Sevcik, principal of Northeast Consulting, a Boston-based technology advisor. "The ATM leader at the moment is Fore Systems, based outside of Pittsburgh," he says.
However, other data communications vendors are seriously considering the chip. For example, Racal-Datacom, in Sunrise, FL, is building the Cubit chip into a family of multiplexer products to be released in 1996.
"These chips may access public or private ATM networks, including those that are fiber-based," says Philip Wilson, the company`s area vice president for engineering. "There are other ATM chipsets on the market," he notes, "but Cubit has an on-chip backplane bus, high bandwidth, flexible interfacing, ATM cell handling and bus-access management all on one chip.
"Racal-Datacom`s ATM systems should integrate well into fiber-optic applications," Wilson adds. "One thing that the ATM Forum standards body did correctly was to adopt the existing Sonet backbone structure. We already offer a Premnet time-division multiplexer that uses the Sonet backbone (see page 6) but Cubit may add full-speed local area network and legacy-computing communications from the user side, in addition to multiple outgoing network connections, including DS-3 and OC-3 capabilities.
"We could also use Cubit to bring fiber and ATM to the desktop," Wilson continues, "but for now, we see ATM primarily as a tool for building backbone networks within buildings or across campuses. It fits the common hub-star paradigm quite well. You can put proprietary and standard local area network traffic over it and do all the things that people promised to do with [fiber distributed data interface]...but didn`t. So I see premises networks as the first major use of ATM. Carriers and competitive access providers also offer ATM services, but their first successes probably will be in [local area network]-speed metropolitan-area connectivity."
Will users buy?
It may be a stretch to think that users will embrace local area network products built around wide-area telecommunications technologies such as ATM and Sonet, according to Sevcik. "Transwitch is talking about a high-performance fiber application here. They`re discussing building Sonet/ATM stackable hubs, but users may prefer to stay with unshielded twisted-pair premises networks."
Still, users are curious about ATM, analysts say. To illustrate, last year the Business Research Group studied the internetworking plans of 400 major U.S. corporations and found strong interest in ATM. "People are evaluating it as a future possibility, even though few real products exist," O`Neill says. "Network administrators seem to see ATM as a backbone to tie together distributed enterprise networks and workgroups."
He cautions that user interest, network services and access products may not guarantee ATM`s market success. "It`s one thing to have ATM capabilities, and quite another to distribute, manage and bill these services cost-effectively," says O`Neill. "So the Business Research Group is about to study whether the regional Bell operating companies, interexchange carriers and competitive access providers are evolving their network- and operational-support systems to effectively handle ATM services. If you can`t monitor it or bill for it, then you can`t really do it, can you?" q
Dave Powell writes from Winchester, MA.