Today, data traffic is eclipsing voice traffic in volume, and demand by both business and consumers for a wide array of interactive services is growing. Moreover, in the new world of the Internet, voice communications are just another form of data, which has profound implications for the future of digital subscriber line (DSL) broadband services.

Interactive voice services will converge with interactive data services at the computer desktop. Telephony will become embedded-along with other services-in computer networking applications. Traditional voice calls, unified messaging, virtual assistants, and audio and videoconferencing, together with application services, Internet access with voice recognition, virtual private networks, intranets, extranets, and gaming and chat with interactive multimedia-based voice and video, will all be accessible with the click of a mouse. The future of the DSL services industry will belong to those broadband providers that can most cost-effectively and profitably deliver this rich panoply of easily accessible, concurrent services.

Converged desktop data and voice services, ranging from "best effort" to high quality of service (QoS), will be transported across communications networks in Internet protocol (IP) packets. The exponential growth of data traffic in an Internet-enabled economy is already driving the next-generation network core toward high-bandwidth, packet-based technologies. As voice moves into packets on integrated access lines, and as converged applications emerge at the desktop, only IP combines proven scalability with convergence support. The present and the future of DSL will belong to providers that prepare now to deliver both voice and data services in an IP format.

That may not seem evident at a time when emergent IP technologies co-exist with established circuit-based technologies-asynchronous transfer mode (ATM) and Sonet -that are still prevalent at the network core. But the economics of IP are compelling and inescapable. Enabling dynamically scalable speeds from 1 megabyte/second to 1 terabyte/s, the high-performance, IP-based technologies proliferating in metropolitan-area and wide-area networks today can move both toll-quality voice and data at a fraction of the cost of ATM and Sonet technologies. The investment community is looking toward the future and has already placed its bet on IP. In 1999, IP-centric networking companies, including Copper Mountain, raised a combined $24 billion in initial public offerings, and today research and development spending for packet technologies far outstrips R&D spending for ATM.

In today's increasingly multimedia-rich networking environment, IP-based DSL access multiplexers (DSLAMs) give DSL service providers an immediate, decisive advantage over providers that rely on ATM-based DSLAMs. That advantage derives from a fundamental difference between IP and ATM: IP technology can distinguish between various services and make per-hop routing, aggregation, traffic shaping and priority decisions, while ATM technology can only distinguish between virtual circuits. That key difference enables providers that standardize on IP-based concentrators to create, provision and scale new services more easily and cost-effectively than providers that rely on ATM-based DSLAMs.

ATM transports voice and data traffic in fixed-size, 53-byte cells. In the ATM world, all cells, and therefore all traffic, look the same, and an ATM-optimized DSLAM cannot discriminate between one kind of traffic and another. IP works with Internet packets, which are larger than ATM cells and distinguishable by type of service and destination. An IP-based concentrator can see these packets in their entirety, and discriminate between different kinds of traffic and between traffic from different subscribers.

Because all traffic looks the same to an ATM-optimized DSLAM, a DSL service provider using ATM DSLAMs must provision separate end-to-end permanent virtual circuits (PVCs) for each kind of traffic, and for each subscriber. This requirement exponentially escalates an ATM-centric provider's scalability and management problems, which are further compounded when the provider attempts to provision voice traffic: With an ATM DSLAM, the provider must set up separate PVCs for every subscriber and for every call by every subscriber.

Take, for example, a DSL service provider that signs up 100 subscribers for a full menu of services, including QoS VoDSL. To provision 10 voice calls for 10 subscribers with an ATM-optimized DSLAM, this provider must set up and manage 100 PVCs. In contrast, a DSL service provider that standardizes on DSLAMs with IP service intelligence can aggregate all voice subscribers into one low-delay path through the entire network. Where the ATM-centric provider must set up 100 PVCs to provision 10 calls for 10 subscribers, the IP-based provider can provision the same number of calls with only one network path.

DSL service providers opting for IP-optimized DSLAMs over ATM-based DSLAMs enjoy another significant advantage: They can deliver voice services without deploying expensive, centralized Class 5 circuit switches. Initial carrier-class VoDSL solutions have relied on GR-303 gateways that use IP packet or AAL2 ATM cell encapsulations for voice traffic. These solutions are still dependent on Class 5 switches. But new, more cost-effective packet-based switching solutions are now available to providers that deploy IP-optimized DSLAMs.

These next-generation switching solutions-which include soft switches and trunk gateways-use advanced signaling protocols like Media Gateway Control Protocol to provide voice routing and calling features that have been traditionally provided by Class 5 switches. Running on standard server hardware or fault-tolerant computers, soft switches provide call control, administration and custom calling features for packet voice network devices. Trunk gateways are central office hardware platforms for internetworking between packet voice networks and the public switched telephone network. Used together as a distributed switching system, soft switches and trunk gateways perform call control and routing for subscriber telephony traffic placed between packet voice endpoints or between packet and circuit-switched subscribers.

Soft switches and trunk gateways take the features and functionality of the GR303 gateway and Class 5 switch and partition them among multiple devices that can be distributed throughout the carrier network at optimum points and properly sized for each point of presence-at much lower cost.

A small number of telecommunications equipment makers have opted to overcome ATM's inherent provisioning and scaling liabilities with development efforts that rely on automatically established switched virtual circuits (SVCs). In a world with no Internet, where the main challenge would be more cost-effective provisioning and scaling of telephony services, this proposed graft on the existing ATM infrastructure would make sense. However, ATM SVCs will do nothing for nonvoice services, which will still move in packets and require an entirely different, IP-routed network architecture. Moreover, by the time an SVC-enhanced ATM access network can be tested and brought online, a New World IP network core capable of handling converged voice and data services will be fully deployed. The "new and improved" ATM SVC solutions will be obsolete by the time they become operational.

While new service architectures must be compatible with existing ATM transport networks, it makes no sense to invest further in service architectures that depend on ATM. Voice service integrated access devices (IADs) and gateways that put voice bitstreams directly into ATM cells are a dead-end solution that will not support unified services. Service providers that go down this road are heading into a box canyon. On the other hand, IADs and gateways that put voice into IP packets provide flexibility. IP packets may be segmented into ATM cells for transmission, but can be reassembled at network nodes for aggregation and routing based on IP header information-or they may be sent as packets without the additional overhead of ATM segmentation.

DSLAMs with IP service intelligence can cost-effectively provision and scale QoS VoDSL across ATM-dominated network cores now, in conjunction with voice gateways and packet-capable IADs. Voice-over-IP (VoIP) solutions are the optimal choice for DSL service providers provisioning voice services today.

Despite its prevalence at today's network core, ATM/Sonet is an Old World technology, a legacy of the 20th-century paradigm of separate voice and data networks. New World converged services networks will be IP-packet-driven. ATM-based DSLAMs cannot migrate to these new networks; DSLAMs with IP service intelligence can. Service providers that anticipate the coming shift from circuits to packets and select the right networking platform now will dominate in the marketplace of the 21st century.