Networks: The next generation

The somewhat hazy notion of a new carrier network is starting to take shape. First came news last week that Sprint bought a 40-channel dense wavelength division multiplexing system that is expected to start carrying traffic within six months. The system will increase Sprint's network capacity, now at 16 channels, by 250%.

Then came word that Bay Networks and Ericsson have agreed to deliver end-to-end IP switching solutions based on the so-called multiprotocol label switching (MPLS) standard by the end of the year.

Both news items raise questions about how 21st century networks are evolving. Tom Nolle, president of CIMI Corp., points out an economic truism that's becoming increasingly obvious to carriers: WDM is an example of technology's relentless influence on the carrier pocketbook. In other words, WDM reduces the cost per wavelength of transport bandwidth.

At the same time, WDM is carriers' sole option for deploying theoretically infinite bandwidth across the backbone, which today's data demands require. New players such as Qwest and Level 3, which offer virtually unlimited bandwidth, are quickly changing what once was a demand-side market into a supply-side market-a situation cited by none other than Kathy Szelag, director of strategy for Lucent's optical networking business.

Another way carriers are driving down costs is through Internet protocol-based services. These connectionless services must be provisioned over a connection-oriented infrastructure to ensure quality of service (QOS) for users and a solid return on bandwidth investment for carriers.

Enter MPLS, touted as the solution to QOS issues. Although not yet a final standard, MPLS decreases latency, increases performance and throughput, and makes a multivendor IP-over-ATM network practical.

Now comes the sticky part. First, ATM's value lies in its ability to conserve a costly resource, i.e. bandwidth. So it's valuable only as long as bandwidth is a scarce commodity.

Nolle estimates the window of opportunity will last 10 years before WDM and other factors pull the unit cost of bandwidth down so low that statistical multiplexing becomes a wasted effort. "It's a delicate issue of market timing," he says.

Probe Research predicts it will take eight years for spending on broadband switching gear-such as ATM and MPLS-based equipment-to outpace circuit-switched spending in the public network. And industry market experts agree that routers and switches eventually will plug directly into the optical layer, eliminating the Sonet multiplexing function.

So the race is on. One of the handicappers is the Internet Engineering Task Force.

The task force must determine final details of MPLS, which shares characteristics of Cisco's proprietary tag switching and IBM's Aris IP switching solution.

Most of the final wrangling is over details, and Cisco says it won't use its considerable influence to sway the task force. Thomas Downey, Cisco's director of product marketing, says Cisco has always expected the MPLS standard to differ from tag switching. "We wanted multiple [vendors] to do one standard," he says. The MPLS standard is expected late this year or in early 1999.

Competitors on the Internet also are handicapping the race by going ahead with their own QOS ideas. One is Savvis Communications Corp., a St. Louis-based national service provider that recently started a new way of connecting national Internet provider networks without going through often-congested public network access points.

The resulting high-quality connections provide a scalable infrastructure with low latency, says Mike Gaddis, the company's executive vice president of engineering and chief technical officer, who helped build one of the first ATM switches.

And in case anyone thinks the most futuristic infrastructure of all-the all-optical network-is a distant dream, keep your antennas tuned to upcoming Supercomm news from the likes of Lucent, Alcatel and Fujitsu. Lucent's Szelag is willing to predict that Al Gore's vision of an all-optical, nationwide backbone network could become reality in 2002. Any handicappers?