PDF of this tutorialCarriers are moving voice services to packet networks both to reduce upfront and operational costs and to provide more value-added services in an increasingly competitive environment. A recent study by a major carrier found that packet equipment was 70 percent less expensive than traditional voice equipment, and data access lines were 60 percent to 80 percent cheaper than voice lines. Maintenance of packet networks was 50 percent less expensive, while provisioning was 72 percent lower. However, consolidation of voice from the PSTN onto packet networks has, in the past, proven difficult and therefore has happened very slowly. International voice-over–IP call volumes, which provide the most compelling business case for packet telephony, are still a drop in the ocean of international telephony traffic but have experienced phenomenal growth since 1998, according to a recent report by Washington, D.C.–based research group TeleGeography. According to the "TeleGeography 2001" report, which contains results of an exclusive survey of major voice-over-packet (VoP) providers in 1999 and 2000, international Internet telephony traffic volumes reached 1.7 billion minutes in 1999—a growth rate of more than 1,000 percent from 1998. IDC projected more than 9 billion minutes of voice traffic to travel over worldwide packet networks in 2000, exceeding 135 billion minutes in 2004. Service revenue is projected at $1.6 billion in 2000 and $18.7 billion in 2004.
While it is clear that VoP is growing, there is still considerable debate about whether the underlying network technology will be ATM or IP. At the edge of the network the choice, driven primarily by the regional Bell operating companies (RBOCs), is ATM. An ATM–dominated access network is clearly in the works because until recently IP did not provide the quality of service (QoS) guarantees that are so important for voice. Although QoS protocols such as DiffServ, resource reservation protocol (RSVP), and MPLS have been implemented, most of today's IP traffic is actually being carried over ATM. However, in the long term with the recent success of MPLS it appears that pure IP over lambda may be the winner. And certainly, IP at the application layer and the desktop is a more than just a viable near-term situation.
In addition to the challenges in architecting networks with end to end QoS, service providers must ensure that the rollout of such networks cause no disruption to their existing voice service revenue, which currently represent about 80 percent of their overall revenue source. With more than $650 billion of worldwide revenue generated by traditional voice and fax services and more than $250 billion installed base of traditional equipment infrastructure in the United States alone, service providers must deploy next-generation packet switches that seamlessly interconnect and competitively function as time division multiplexing (TDM)–based PSTN switches as well as support voice over ATM (VoATM) and voice over IP (VoIP).
It appears that most carriers, especially the larger incumbent carriers will start the migration to packet telephony on the trunk side first (Class-4 tandem) and eventually migrate to the access (Class 5). This migration model is similar to the migration from analog switches to digital switches, which started in the late 1970s. Carriers first started on the inner network (i.e., tandem) and then moved outwards to the Class 5.
The architecture for VoP, the reasons for choosing IP or ATM, and considerations in next-generation system design need to be understood to accelerate VoP deployments.
