Due to the overambitious plans of alternative carriers, many of these broadband deployment plans remain on the drawing boards, as the carriers have gone under in the rough climate of 2001. But access networks have another bottleneck to solve besides that of simply getting broadband service delivered to the end user.

The proponents of Internet Protocol must tout the advantage of all-IP access to an audience comfortable with dial-up, circuit-switched service. Even those business customers familiar with broadband services tend to gain those services through leased-line T1/T3 networks, which retain a circuit-switched base. If we analyze the incoming access traffic from the core of the network, the trends toward "IP-ization" at the edge are scarcely encouraging.

The best assessment of the amount of voice traffic traveling as IP packets in the global network centers on a number between 8 and 9 percent. It is true that, in a horrendous year for all types of equipment sales, voice-over-IP gateways have fared better than most other IP switching and routing product categories. Sales have stayed relatively flat over the last few quarters, averaging just under $250 million per quarter, or less than $1 billion for the year, according to Cahners In-Stat. That’s compared with quarterly downturns of between 10 and 30 percent in 2001 for most switching and access product categories.

But the push is not coming from the users of PBXes and computer telephony equipment, since IP-capable PBXes still amount to scarcely more than 1 percent of all PBX sales. The advantages of packets are promoted by carriers, who can only show advantages in the cost of transport, not in the services for the end user. And what is true for existing carriers with a circuit-switched infrastructure is even truer for carriers trying to enter voice telephony through alternative physical-layer transport.

The cable TV industry relied upon CableLabs to provide voice prioritization in the Docsis 1.1 cable-modem standard, and better channelization in Docsis 2.0, under the assumption that many set-top boxes would be offering IP telephony services over coaxial cable by now. While larger multiple system operators (MSOs) like AT&T and Cox Communications Inc. are continuing some interesting regional trials, the penetration of MSOs into telephony carriers’ voice business has been scarcely measurable, even in single digits.

"Operators, particularly in a year like this one, are not going to adopt packetized voice simply because it’s the next new thing," said Richard Prodan, chief scientist with cable infrastructure specialist Terayon Communication Systems Inc. "When calls become cheaper to make in the aggregate than circuit-switched phone calls, you’ll see some operators really move to voice services–but not before."

The issue of what to do with time-division multiplexing traffic multiplies as packets leave the home or enterprise and enter the first access mile to the metropolitan ring or trunk line. For cable TV networks, Ethernet remains the flavor of the month–CableLabs decided long ago to take advantage of Ethernet cards in the home computer by opting for a native Ethernet framing interface. DSL networks, by contrast, use an asynchronous transfer mode interface at the access point, though more and more DSL modem vendors have been adding Point-to-Point Protocol over Ethernet connectivity on top of ATM.

The issue becomes testier as Ethernet makes greater forays into carrier networks. The arrival of Gigabit and 10-Gbit Ethernet made the framing method appropriate for aggregated packet transport in the wide-area network (WAN). In 2001, the formation of the Ethernet in the First Mile study group within the IEEE, as well as the Metropolitan Ethernet Forum, enhanced the viability of using Ethernet frames, perhaps in conjunction with passive optical network connections, as an access method for all traffic leaving an enterprise or home for the WAN.

Notice that no particular dispensation is made for circuit-switched traffic in this model. Companies that tout high-speed Ethernet connections in the last mile are betting that quality-of-service (QoS) methods, such as using generalized multiprotocol label switching (G-MPLS) to set up IP flows, will provide users with all the voice and video services they need over IP. Once again, however, market realities may suggest a different evolutionary path as traffic enters a metropolitan ring. Cisco Systems Inc. is one of several companies promoting conversion to IP telephony in small offices and homes, though customers still need to be convinced of why analog voice in a circuit-switched network cannot serve their needs for a long time to come.

There may possibly be a near-term driver for IP service coming from users’ desires for security, though that will accelerate the question of whether G-MPLS can scale as a QoS mechanism for the backbone. Virtual private networks (VPNs), the tunneled technology developed for linking telecommuters securely with a home office, are booming in popularity since the problems of Code Red and other viruses arose in the summer. If there was a key topic at the fall Networld+Interop conference, it was multiplayer security with an emphasis on VPNs, a trend that only increased in the atmosphere following the Sept. 11 terrorist attacks.

The Internet Engineering Task Force has been trying to encourage VPN creation using MPLS, through a standard called RFC 2547, using the Border Gateway Protocol in the core of the network to propagate MPLS VPN information to the edge. Eric Peterson, director of IP Protocol Services at Unisphere Networks Inc., said that RFC 2547 VPN scaling may become a bottleneck in access networks even faster than any problems of using MPLS for voice prioritization.

If MPLS for VPNs can scale, the added security of moving to IP networks adds another reason for businesses and homes to convert all outbound traffic to packets. But it remains a difficult sell to tell customers to displace circuit-switched access equipment in an economic downturn as severe as that seen in 2001.