ATM-based infrastructure to pull fiber to the desktop
Data, voice and video traffic pouring through corporate networks at very high speeds via asynchronous transfer mode technology is expected to drive fiber directly to the computer
gary kim
Workgroup and departmental networking requirements alone are likely to prove insufficient to drive bandwidth needs--and optical fiber--to the desktop. Instead, demand for asynchronous transfer mode technology--and, therefore, fiber to the desk--is predicted to evolve when multiple networks can be collapsed into a single, unified architecture.
In most businesses, the force pushing fiber to the desktop is its ability to collapse multiple networks into an integrated transmission system and to place multimedia applications onto a single access platform. For example, Atlanta`s Northern Telecom has upgraded its internal networks, connecting more than 150 locations and 59,000 employees in 110 countries, through ATM technology and equipment.
The company discovered that when voice and video traffic ranged from 5% to 25% of total traffic, monthly bandwidth costs dropped radically. Whereas typical spending for digital carrier system, level one (1.554-megabit-per-second) traffic exceeded $50,000 a month per node, the ATM transport infrastructure costs $40,000 a month for each node on the network.
"The average corporation has 5.6 networks," says William Conner, Northern Telecom`s executive vice president, multimedia communications systems sales and marketing for enterprise network solutions, "and these application-specific networks typically focus on voice, data, video and training."
Exceptions to the unified network exceptions are high-bandwidth-intensive applications such as computer-aided design and manufacturing, which can justify optical fiber connectivity even at the workgroup level. Hewlett-Packard Co., for example, typically pushes information traffic over its internal networks on the order of 4 trillion to 5 trillion bytes per month. Still, the typical enterprise will see the greatest return on ATM investment when heterogeneous types of traffic are carried over the network.
"Customers see ATM as an elegant, simple network, providing more manageable performance and easier handling of moves, adds and changes," cites Northern Telecom`s Al Delorenzi, Magellan marketing vice president. "The first application that makes sense is in the backbone network, to support server-to-server communications."
Strategic vision is crucial
But operating efficiencies derived from integrating transport might still prove insufficient to move information technology executives toward ATM and optical-fiber access in the workplace. Greater concerns, such as increasing revenues, increasing customer satisfaction and enhancing teamwork, are crucial executive motivations, says Conner. Companies implementing ATM technology, for example, offer a competitive edge, on the order of "30% lower cost and five times the success," he says.
More important, though, is a company`s clear strategic vision of why the information technology investment is made and why ATM and fiber to the desktop are required. The 1970s model of information technology, based on mainframe computing, stable markets and low competition, was a "war of position" information-handling strategy. In that host-computer-centered scenario, maneuverability of company information was difficult.
In the fast-paced markets of the 1990s, which are based on client/server technology, massive competition and time-based advantage, a "war of maneuver" information-handling strategy is paramount to a company`s success. In this new scenario, decentralized computing should not only transform business operations, but should also allow a company to create marketplace dislocations and then structure the platform for exploiting those dislocations.
Consolidating traffic using ATM is more than process reengineering; it is a means to a strategic goal--organizational nimbleness. The objective is not just to improve organizational effectiveness, but to create opportunities for growth. "War of maneuver" marketing is suited to defending a company`s market share. "War of position" marketing seeks to create new markets and share existing market activities.
In the meantime, of course, there will be no short cut to ATM technology. In part, that`s because few enterprises are as bandwidth-enabled as Sun Microsystems Inc., Hewlett-Packard Co. and other workstation suppliers. These companies pioneered the concept of distributing computing operations using local area networks.
In the late 1980s, teams working on the design of reduced instruction set computers, for example, saw networking bottlenecks on their contention-based Ethernet LANs. So they moved to switched designs that clustered groups of workstations on hubs. The result was to boost the throughput for individual links to the rated 10-Mbit/sec throughput of Ethernet LANs in place of the 1- to 2-Mbit/sec rates that many users actually accessed on their contention-access networks.
The hub strategy has led to upgrading twisted-pair copper-wire Ethernet networks to 100-Mbit/sec performance. In large part, this interim strategy would preserve the existing infrastructure of 802.3-standard network interfaces and management.
Today, two major copper-network proposals are under development: The Fast Ethernet Alliance and an Institute of Electrical and Electronics Engineers 802.3 standards group are working on a 100Base-T standard. This approach aims to preserve the use of copper media. Among companies supporting 100Base-T standards are networking suppliers Cabletron Systems, Chipcom Corp., Intel, 3Com and Bay Networks (formed by the merger of Wellfleet Communications and Synoptics Communications). The competing 100VG-AnyLAN standard is backed by Hewlett-Packard Co., AT&T Network Systems and IBM.
There are, of course, some network distance limitations to each proposal. The 100Base-T network is limited to a diameter of approximately 250 meters, almost one-tenth the maximum length of an IEEE-802.3-based LAN. The 100VG-AnyLAN network is limited to a 4000-meter diameter. Whereas the 100Base-T network retains contention-access methods, the 100VG-AnyLAN network substitutes a more-deterministic access method to avoid data collisions. So far, the two proposals are incompatible, suggesting that major pricing moves for ATM nodes, compared to either of the 100-Mbit/sec IEEE-802.3 networks, could be important in building customer support for any of these options.
Cost factors
The primary hurdle for ATM proponents is cost. Few IEEE-802.3 or Ethernet sites have installed fiber-optic cables, so major rewiring is required before the shift to ATM can be made. There also are moves backed by IBM to develop a 25-Mbit/sec access scheme for ATM. These moves would allow the use of Category 3 twisted-pair copper wiring and the reuse of the copper infrastructure already in place.
Other network technologies, including fiber distributed data interface and copper distributed data interface, can support 100-Mbit/sec transmission over distances to 2 kilometers. Though pricing for network interfaces continues to evolve, FDDI/CDDI costs per node are said to lie between those of Fast Ethernet and ATM.
It may be some time before ATM technology reaches the desktop and supplants Ethernet and IEEE-802.3-standard connections. And even when it does, fiber to the desktop might not follow. "Even if you use ATM to the desktop, that doesn`t necessarily require installation of new fiber," says Peter Alexander, Bay Networks` director of enterprise networks. "IBM, for example, is pushing 25-Mbits/sec ATM over copper." He claims that ultimately, "raw speed and cost will drive fiber" to the desk. The time is coming when end users will need to access "a zillion gigabits," says Alexander. Also, the cost of fiber is dropping, especially in new buildings, so it makes sense to install it.
Still, applications will drive everything. Multimedia and videophone calls made off the LAN are some drivers for fiber, Alexander says. Already, LAN transfers of graphics, such as presentations, are slowing down networks. "We run into this every day," he says. Faster central processing units, falling memory prices and client/server architectures, which necessarily generate traffic, are other bandwidth drivers. "As the central processing units get faster, files get larger. And as memory costs go down, users don`t have to compress files, which again drives up bandwidth requirements," he adds.
When fiber-to-the-desktop networks are put into place, a new type of management system will be required to handle the inevitable moves, adds and changes that all LAN administrators must cope with, says Ken Garrett, Astarte Networks vice president for sales and marketing. In a campus environment, for example, a backbone fiber network might fan out to 100 to 300 desks, he says. In previous networks, an automated and remote ability to reconfigure the physical layer facilities might not have been essential. But fiber to the desk is entirely different and creates a need for a different kind of facilities management, Garrett adds.
The advent of ATM technology will cause fundamental rethinking about network function and design. For example, it is less a LAN data technology and more a fundamental communications channel. Whereas LAN traffic is not highly predictable in the delay and throughput areas, ATM is highly predictable. This characteristic suits the support of such ATM applications as voice and video, neither of which are tolerant of packet delay and arrival times.
Renewed interest has, therefore, emerged in computer/telephony integration--the ability to support voice and personal computer communications over one network. In approximately four years, the computer/telephony integration market may represent nearly $8 billion in annual revenues in hardware, software, consulting and system integration activities in North America alone, according to Northern Telecom executives.
The high-bandwidth capacity of fiber to the desk becomes important because advanced computer/telephony integration includes voice and low-speed data traffic, high-speed data transfers to accommodate desktop videoconferencing and support for Microsoft Corp.`s Windows 95 operating system.
In the background and regardless of end-user applications, the raw horsepower of computing devices is growing exponentially, creating additional requirements for bandwidth at the desk in workgroup, enterprise and public and private networks.
Even as ATM-based networking gradually supplants the installed base of Ethernet nodes, network solutions such as Ethernet switching and Fast Ethernet, operating at 100 Mbits/sec, are projected to gain ground.
Opportunities in the networking market are enormous. Personal computer LANs supported 18.5 million nodes in 1994 and might achieve 23.2 million nodes in 1995, according to Fleck Research, a Santa Ana, CA-based company that studies connector markets.
International Data Corp. estimates a base of PCs connected to LANs in the 40 million to 55 million range. u
Gary Kim is senior vice president of Probe Research Inc. in Cedar Knolls, NJ.