April 1998

BY JEFF HOLTMEIER

In the case of xDSL, the letter "x" highlights a lack of uniformity, and serves to emphasize the challenges faced by Digital Subscriber Line technology as it competes against cable, satellite, and wireless innovations for the next generation of Internet access. However, despite the many different flavors of technologies that are grouped under the xDSL nameplate, there are several important efforts underway to build a stable, interoperable set of products that can deliver the promise of megabit access over the existing telephone network.

THE POTENTIAL
According to the ADSL Forum, there are more than 560 million copper loops worldwide that are candidates for some form of DSL. Much of the required infrastructure is already in place, including the copper plant, central office and curbside concentration facilities, call tracking and back-office accounting systems, and existing customer relationships. There is clearly demand for high-speed Internet access, both from commercial and residential customers. Internet content continues to grow richer and more bandwidth-intensive and, despite fears that the Internet might collapse due to explosive bandwidth demand, backbone providers are continuing to expand capacity and groom their services to keep up with additional requirements. For example, MCI recently upgraded a section of its core network to 80 gigabits per second, enough capacity for half a million simultaneous phone calls.

THE CHALLENGE
So what’s the holdup? To succeed, the industry must address several important issues, most critically, the challenge to develop open, stable interoperability standards so that competition and volume can drive down the cost of installing and operating DSL circuits. Complicating this evolution is the fact that several groups have their hands in the development of the standards. For example, physical layer issues, like line codes, have been the province of the American National Standards Institute (ANSI), European Telecommunications Standards Institute (ETSI), and the International Telecommunications Union (ITU). The adaptation of higher-layer protocols like Asynchronous Transfer Mode (ATM) and the protocols of the Internet to the subtleties of DSL have fallen mainly to the technical working groups of the ADSL Forum. Yet, despite the work of these groups, there are still many competing standards and very few interoperable implementations. Some vendors that have announced interoperability are not yet compliant with the standards established by either ANSI or the ADSL Forum.

To demonstrate the breadth of the problem, Table 1 describes some of the technologies that are typically classified as DSL. None of these technologies will communicate with any other.

But just agreeing on one of these technologies is not sufficient to guarantee interoperability. For example, DMT ADSL has been chosen as a standard by ANSI (see ANSI T1.413). There are several vendors who currently produce or have announced DMT-compliant chipsets, but, at this writing, there has not been an interoperability demonstration between any of these vendors.

Even equipment vendors that use the same manufacturer’s chipset have some major hurdles to overcome before their equipment can transfer information over a copper loop. For example, the ANSI DMT specification includes several different upstream and downstream channels and each of those channels can optionally include an extra error reduction step called "interleaving." Without agreement on these and other parameters associated with DMT, there can be no communication across a DMT line.

Once agreements on chip manufacturer and chip configuration are in place, the equipment vendors must agree on how to interpret the bits that are flowing across the DSL line. Many early implementations of DSL were focused on delivering video-on-demand to the home and used a video streaming protocol. Today’s focus on data and voice has brought traditional LAN and WAN protocols to the DSL arena. For most vendors and potential DSL customers, Internet Protocol (IP) is the highest priority. While there are standards for carrying IP over Ethernet or Token Ring in a LAN environment, and for transporting IP over ATM, Frame Relay, or PPP in a WAN environment, all of those standards require some modification to work with DSL. Vendors have invented a variety of variable-length framing techniques, including derivatives of Ethernet, PPP, and Frame Relay. The fixed-length ATM cell camp has also tossed in several alternatives, including Frame UNI, Cells in Frames, and PPP over ATM variants. The technical working groups of the ADSL Forum are working on clarifying these alternatives, with partial standards already in place for ATM and Frame Relay transport. The Internet Engineering Task Force (IETF) has also recently commissioned a working group for DSL technologies.

Once the vendors have agreed on line code, chip configuration, and data transport, management and reporting protocols must be in place. DMT ADSL has an embedded operations channel that was designed for simple status and configuration traffic. Richer alternatives have been proposed that use existing standards such as SNMP. The issue of in-band versus out-of-band management is not yet settled, and in fact, there are proposals to use both technologies at different levels of the protocol stacks.

Many DSL equipment vendors have avoided the multivendor interoperability issues altogether by developing equipment for both the customer premise, and the telco office ends of a DSL line. While this will provide a working implementation, it constrains the feature set available to the customer and raises the typical concerns of having a single supplier. For example, a vendor might only provide a card that installs in the PCI slot of a PC, ruling out ADSL for notebook owners. For broad customer acceptance of DSL technologies, solutions must exist for single desktop PCs, notebooks, homes with one or multiple PCs, and small to medium-sized businesses. There are opportunities for DSL customer premise equipment (CPE) in a variety of forms:

• Single user, internal PC NIC cards.
• Single user PCMCIA notebook modems.
• External modems with 10BaseT Ethernet interfaces.
• External modems with USB (Universal Serial Bus) interfaces.
• Routers with integrated DSL modems.

Currently, no DSL equipment vendor provides this broad a range of products. Interoperability standards would allow vendors to specialize in a few of these areas and partner for others. Customers would be assured of a broader range of solutions.

The hurdles to developing a healthy multivendor market for DSL include agreement on line codes, chip configuration, and transport and management protocols. Chip vendors must follow through on their interoperability pledges. The standards bodies must continue to eliminate ambiguities in existing standards and extend the standards to cover all the issues that affect compatibility. Adherence to these developing standards will require vendors to abandon proprietary characteristics of their system and the perceived market advantages of those features in order to enjoy the benefits of a larger market.

DSL’S COMPETITION
These interoperability issues might not be a concern if the technology could ripen slowly. However, there are several very attractive alternatives to DSL that are being deployed as competition for broadband Internet access. Cable, satellite, and wireless are all capable of delivering similar features to the same customers at a comparable cost.

The major cable companies in the United States, including Time Warner, MediaOne, Cox Communications, TCI, and ComCast, are all deploying cable modems. Internet delivery over cable has several significant hurdles, including:

• Upgrading the cable system to handle two-way communications.
• Security and reliability concerns associated with a shared cable (compared to the dedicated bandwidth provided by DSL).
• Limited access to high-margin business customers.

However, the cable industry bypassed the interoperability challenge that’s facing the DSL industry. It used CableLabs, a research and development consortium funded solely by cable television system operators, to develop and disseminate standards. By excluding equipment manufacturers from the process, standards evolved without the problems of individual company agendas. The cable operators are enforcing the new Data Over Cable Service Interface Specification (DOCSIS) standards by requiring compliance from their vendors.

The technology behind the successful 18inch Digital Satellite System dishes provides a nationwide infrastructure for delivering Internet content at up to 0.4 MB per second. For some areas, satellite is the only broadband alternative. Satellite suffers from its one-way architecture; while Internet content is beamed from satellite to dish to PC at kilobit speeds, messages from the customer’s PC must travel to the Internet over slower analog modems.

High-bandwidth wireless technologies are finally approaching cost parity with these other technologies. Last June, CellularVision rolled out 0.5 MB/second wireless service based on Local Multipoint Distribution Services (LMDS) in Manhattan. The initial offering has the same limitations as one-way cable and satellite — the upstream path is through a regular analog 28.8 modem.

Of these alternatives, two-way cable systems are clearly the most direct threat to DSL. The CableLabs consortium is actively addressing the security issue, cable providers are ramping up their infrastructure to support the Internet, and two-way communications and standards are driving down the cost of the supporting technologies. Monthly fees from MediaOne in Boston are currently under $50 for both cable access and ISP fees, significantly lower than the $160 that a customer will pay to PacBell and a service provider for FasTrak DSL service in the San Francisco Bay Area.

THE SOLUTION
Is it curtains for DSL? Absolutely not. As the DSL equipment vendors have moved from research and development to deployment, customers and competitors assert that there is a lot to be gained by working together, but also warn that a fragmented market will keep costs prohibitively high. Fortunately, the concerns expressed by telcos and CLECs (competitive local exchange carriers) are compelling equipment vendors to work towards achieving interoperability within the realm of DSL.

Interoperability is developing along three axes: bilateral cooperation, maturing work in the standards bodies, and active multivendor working groups. Examples of bilateral cooperation have been in place for a year or more. Westell and DSC have cooperated on a project for Bell Atlantic. More recently, 3Com and AFC, and Paradyne and AG Communications, have announced interoperability between their CPE and Central Office gear.

The T1E1.4 committee within ANSI has been working on Issue 2 of the DMT specification and a working group has been writing a draft standard for CAP and related line codes. In the ADSL Forum, several Technical Reports (Forum terminology for standards documents) have been finalized and published over the last year. Others are nearing completion, including documents on wiring at the customer premise, network management, testing, and operations and maintenance. Existing standards are being enhanced as customer applications become clearer. At the December 1997 meeting of the ADSL Forum, more than 40 companies supported a contribution, in its third revision, detailing an end-to-end architecture based on PPP over ATM.

Very recently, several multivendor initiatives have sprung up. Westell, one of the early leaders in the DSL market, is organizing two interoperability workshops in 1998, one for CAP and one for DMT, to promote compatibility at all the different layers of the protocol stack. And in January, Microsoft, Compaq, Intel, and a host of DSL vendors and telephone companies put their weight behind a new Universal ADSL Working Group (UAWG) effort to develop standards for the splitter-less version of DMT ADSL. The UAWG’s goal is to ensure that Universal ADSL is the preferred modem technology by the year 2000.

The bandwidth limitations of current modem technology date back to the invention of the telephone. Research conducted more than a decade ago created a capability to shatter that ceiling in the form of Digital Subscriber Line signal processing. Many different versions have been developed to address applications that vary in speed, symmetry, and reach, and high-speed Internet access has created significant market demand for the capabilities. Now, the technology must reach the next step before mass deployment is possible. This includes the elimination of incompatibilities across the technology, and the development and enforcement of standards. Dedication to these goals will enable customers to reap the benefits of innovation at an attractive cost, particularly when compared with competing technologies. When interoperability finally becomes a reality, xDSL will reach its full potential.

John Stephens is director of development and chief technology officer at Cayman Systems, Inc., a leading provider of customer premise equipment for all the major DSL technologies. He has been contributing to DSL standards as editor of the Packet Mode Working Group of the ADSL Forum for the last 18 months. For more information, contact Cayman Systems at 7812791101 or visit the company’s Web site at www.cayman.com.  Email to the author ([email protected])  is also welcome.