Getting from circuit to packet without losing quality: Providers must begin now to prepare their networks to meet the challenges of the data-driven future - Focus on: Voice over Broadband

Telecommunications, Feb, 2002 by San-Qi Li, Stephen Barnes, Jingdong Ye

Carriers worldwide are reshaping their long-held ideas about network architecture. Traditional TDM-based circuit switching is highly capable of supporting a variety of services including toll free dialing, calling cards, advanced caller ID and switched data calls for video conferencing and dial-up Internet access. The rapid growth of Internet traffic, however, and the proliferation of innovative services that integrate traditional voice with information-age features (e.g., 3G wireless, instant messaging, broadband access) are far exceeding the capacity and capabilities of the current infrastructure. Faced with pressure to find new sources of revenue and increase profits, network providers are accelerating plans to deploy packet-centric voice switches that can achieve both goals.

Switch Tectonics

Two key challenges exist:

* The need to create new revenue streams by adding features to traditional services or creating entirely new products;

* The need for network optimization (or modernization) to reduce operating costs and increase margins. Carriers must create networks capable of generating voice and data revenues, while maintaining quality and reducing capital and operating costs.

Traditional TDM circuit switching is converging with emerging packet-switched models, and consumers now want services that meld traditional circuit-switched products with data-enabled services, Although data traffic is growing at a far quicker pace than voice (an estimated 300 percent increase per year for data compared to less than six percent for voice traffic), revenues from voice services continue to represent the majority of carrier's income. As carriers expand their data network capacity, it is only natural that they will merge the circuit-switched voice onto the larger capacity data networks at some point, as long as it is cost effective and does not compromise quality or reliability.

As with the voice network, data network architectures devised in the 1970s and 1980s (e.g., IP and frame relay) were not constructed to support services such as broadband data, voice and video. In fact, of all of the data protocols currently deployed, only ATM can blend various traffic types effectively, while maintaining carrier-grade QoS. Unfortunately ATM lacks the universal connectivity and penetration enjoyed by IP.

IP has shown moderate success in supporting voice and video traffic, but IP-based networks require either gross overbuilding of capacity (usually five to six times the required amount) or complete isolation of voice from data traffic, which does not achieve voice and data integration. With quality extensions, overbuilding capacity can work, but this technique is expensive in the wide area and typically infeasible in the access network, where cost sensitivity is much greater. In addition, network reliability continues to be a major roadblock to quality services over IP.

MPLS (multiprotocol label switching) is gaining support as the final standard for voice, video and data services integration over WANs, but most carriers expect that first-gen MPLS networks won't be deployed until late 2002 or 2003 and that they will be very limited in their ability to support blended services, Second-gen MPLS technology, to be introduced in 2004 and later should have the quality controls required for unified voice and data service support.

This convergence of voice and data, often called circuit-to-packet migration, has proven to be a formidable task. The problems stem primarily from the fact that legacy voice switching equipment is not easily adapted for transmission over data networks, and that data networking equipment (with the exception of ATM) does not have the quality control mechanisms needed to guarantee toll quality voice over either the local loop or the WAN. Changes to both technologies must be implemented to enable voice services on data networks.

Recent attempts to enable voice transmission over data networks using the legacy voice switches have yielded little success. The strategy employed in these early trials focused mainly on adding new voice gateways to existing switches, but the cost, inefficiencies, and complexity of bolting these peripheral systems onto traditional circuit-based switches have made that approach unworkable. Legacy switches are now viewed by many carriers as the bottleneck within their networks, Most carriers have rejected the "half-gen" approaches of legacy add-ons and of dead-end point solutions in favor of a new, more comprehensive approach to the network switching architecture.

With the advent of so many data-centric services (e.g., the Internet, mobile services, enhanced- and instant-messaging, follow-me capabilities, PDAs), the telecom industry now seeks nothing less than a quantum shift in switching technology. Like many upheavals in technology, this trend is driven by the need to reduce power requirements, increase density and boost the carrier's ability to deliver popular, high-margin services.

A Unified Solution