When IP met Sonet

If 1997 was the year of ADSL and other access technologies, 1998 stands to be the year of the backbone.

Fiber optic technology is booming at all levels, from new advances in the fabrication of fiber itself to rapid commercialization of optical transmission technologies such as dense wave division multiplexing (DWDM) and Sonet OC-192.

Here is just a short list of developments in the last few months that have come in response to an anticipated upsurge in large-scale backbone networking:

* In October, Lucent Technologies announced a $350 million investment in its fiber optic operations in Norcross, Ga.

* In Montreal, Northern Telecom has begun volume production of its new OC-192 transmission systems.

* NEC has a stake in several major fiber optic expansion projects worldwide and has made its work with Sprint and AT&T a priority in its U.S. business plan.

But the renewed interest in optical technology is only part of the backbone story. The scramble is on to make sure the global information infrastructure-which a few years ago faced a surfeit of capacity-has enough room to handle the onslaught of Internet traffic that is already building exponentially.

"The data rate for the user interface might soon be 10 Mb/s," suggests Seijiro Yokoyama, senior executive vice president of NEC. By the early 21st century, backbone data speeds could be measured in terabits, he adds.

The rapid shift in network traffic from circuit-switched, connection-oriented architectures to packet-switched, connectionless platforms was predicted. But the speed, utility and great economies of scale of the Internet brought in to this equation were not. People have been talking about ATM-based virtual private networks for years. But it wasn't until now that the Internet became a viable-and desirable foundation for them.

In between the vision and the reality, carriers have been deploying large numbers of IP routers in their networks. So now, like it or not, any would-be backbone has to incorporate scalable IP routers, preferably with optical interfaces to accommodate the necessary throughput. When billing now depends on the number of packets you move, as opposed to the number of minutes a circuit stays open, high-speed routing is everything.

Or as Nattu "Srini" Srinivasan, a technical director at Lucent's Bell Labs, puts it, "Carriers need to get their data traffic to the highest rate possible, as soon as possible, for as long as possible."

Next year, Lucent will set its sights on Cisco Systems' dominance in the IP routing market by introducing a scalable gigabit router designed to bridge current IP networks into the superscaled data networks that will be required to handle the projected Internet growth.

But this goes beyond mere head-to-head competition. Keep in mind that in its latest reorganization, Lucent grouped its optical network and data networking operations under common management.

In doing so, Lucent jumped to the front of the pack in terms of articulating how IP, Sonet and WDM might work together. Alcatel, Nortel and NEC all have strong optical messages. Cisco's 12000 is indeed a carrier-class router. But at this point, no one's connected the dots between the IP technology and fiber optics quite as well as Lucent has.

After that, Lucent's strategy is hardly surprising: Integrate everything. Many of the issues in IP and Sonet are similar: efficient use of bandwidth, network management, compatibility of interfaces and so on. The only difference is that IP uses electrons, and Sonet and DWDM involve photons. What Lucent hopes to create is a platform that combines IP, Sonet and WDM interfaces-and no doubt others-into one large, scalable package that can work in a massive multiprotocol network environment.

"The optical/electronic interface is the most expensive," says Harry Bosco, chief operating officer of Lucent's Optical Networking Group. Enter, then, the high-speed optical cross-connect switch.

Before we get carried away, however, it is prudent to remember that no one really has any sort of optical switch beyond prototype stage. But with the pressure on capacity as it is, expect vendors to begin detailing more of their plans in this regard by mid- to late 1998.

"The ATM and circuit world will come together at this level. It will become cost-effective to move the core switch into an optical cross connect," says Bosco. "We don't see any reason why we can't do it."