AT&T Labs cookin' up a voice and data mix

Projects aim to mix traditional phone service with packet-switched networks.

By Sandra Gittlen
Network World Fusion, 12/17/98

In the Labs Editor's Note: With this report, Online Senior Reporter Sandra Gittlen begins a series of articles on the nation's networking research labs. Over the next few months, she'll take you on a virtual tour of these facilities to show you what's cooking - and what you can expect to one day see on your network.

Florham Park, N.J. - On the walls of AT&T Labs' hallways are old photographs of the first trans-Atlantic phone call, five generations of Alexander Graham Bell's family and other reminders of AT&T's illustrious history.

But behind every door is a researcher racing to figure out how to move the telephone giant from the circuit-switched world into a packet-switched one.

For all their efforts, though, even Larry Rabiner, AT&T Labs vice president of research, says the migration from circuit-switched to packet-switched nets is years away.

"We can't just abandon the plain old telephone system for IP," Rabiner says. "There first has to be a merging of the two networks."

Rabiner has charged his staff of more than 2,000 researchers with figuring out a way to protect the company's century-long investment in the traditional phone network while at the same time taking advantage of the benefits of packet-switched networks, such as lower equipment costs and no regulatory fees.

Profile: AT&T Labs

Director:
Larry Rabiner, VP of research.

Located:
Florham Park, N.J. (also has offices in Silicon Valley).

Hot projects:
Telephony over Packet Switching, which lets enterprise users call one another using simple voice commands instead of phone numbers; the Wireless Integrated Services Protocol project, which lets users move workstations without reconfiguring desktops; the Watson speech recognition project; DjVu image compression; and Cospace, which hooks Web pages to virual worlds and lets Web site visitors view supplementary 3-D content and chat with one another.

Employees: More than 2,000 researchers.

Fun fact:
The phone system that lets a user say "collect" or "calling card" and then processes a person's response was developed by Rabiner and a team of researchers at AT&T Labs.

Two of the hottest research projects underway at AT&T Labs' location here - Telephony over Packet Switching (TOPS) and Wireless Integrated Services Protocol (WISP) - are designed to exploit packet-switching benefits.

Starting at TOPS

TOPS defines a network architecture that lets enterprise users contact one another without having to figure out each others' phone numbers first. Instead, people can use simple voice commands, such as "Call Chuck."

With TOPS, phones would be connected over a packet network to a directory server that houses profiles for each user and uses AT&T Labs' speech recognition software to understand voice commands. The server could be at a corporate site or remotely managed by a service provider from its network. User profiles contained in the directory would include cellular, work and home telephone numbers, along with information on the best times to reach users and the best methods by which to reach them. Profiles could be changed on the fly by network administrators or users.

Information on the party being called would be sent to the caller's phone. If the party being called is not on the packet network at the time, TOPS would connect the call over a public switched telephone network via a gateway.

TOPS, which was developed at AT&T Labs by a team of researchers, including Cormac Sreenan and Partho Mishra, will not be fully implemented for four to five years. However, Mishra says pieces of the project have been adapted for current AT&T services.

WISPing away

Another team collaboration, WISP, combines voice and data across wireless LANs.

Bluetooth, a similar project supported by Ericsson, IBM, Intel, Nokia and Toshiba, lets users connect devices, such as laptop computers, mobile phones and printers, without the use of cables. WISP takes this a step further by adding voice calls to the mix.

With WISP, users would be able to relocate their phones and PCs anytime without having to fuss with rewiring and reconfigurations.

The WISP network prototype features a modified handheld device, which is outfitted with a telephone receiver, a wireless LAN interface card and a 48M-byte Linux-based flashcard, hooked via radio waves to an Ethernet base station located nearby.

The base station, or hub, is connected to the LAN over an ATM backbone. A building could have base stations peppered around its floors. If a user moves out of one base station's geographic range, another base will pick up the handheld's signal and become the user's master base station.

Developing the technology was tricky, Mishra says, because voice has a higher quality-ofservice (QoS) requirement than data and, therefore, its packet delivery needs to be scheduled. With WISP, voice calls reserve the bandwidth they need when the call is established. The base station acts as the arbiter of traffic, deciding which packets get priority.

Mishra says the project works well in-house, but moving it outdoors could be challenging. "There isn't as much bandwidth available as the technology needs," he says. WISP is about five years from being fully implemented, Mishra says.

Although TOPS and WISP still have a ways to go, Rabiner says the migration to combined voice and data networks will happen sooner.

"We need to create intelligent services based on telephony," he says. "But the services available today are a world apart from each other. We need universal broadband access from any device. Remember, there are 290 million phones in use today and they aren't going anywhere."