Communications: Kingston Duffie
Phone lines, down and dirty

Returning to one's roots has been a popular path to rein-vention since Henry David Thoreau surrendered his middle-class sensibilities to the primitive charms of Walden Pond. But when Kingston Duffie left the high-flying world of multilayer switching to reexamine physical provisioning, he found that many in the asynchronous-transfer mode and Internet Protocol switching worlds shuddered at the mere mention of systems reliant on electromechanical parts.

"When we talked to some folks about putting together" 1998 startup Turnstone Systems Inc., Duffie said, "you could almost see them doing the 'Ewww!' routine under their breaths."

But for Duffie, a solution for accurately tracking and documenting the condition of telephone companies' copper pairs was a notion he had carried with him since his first summer job after college. Working in Ottawa as an outside-plant engineer for Bell Canada, Duffie, on his own initiative, took the time and trouble to track the number of copper pairs that would be required to support basic and advanced phone services in new neighborhoods.

"Inevitably, the information would get back to some gray-haired guy who would toss away all our records and say, 'Oh, just put in 200 lines,'" Duffie said. "This was not unusual, which is why so many phone companies today lament their lousy documentation."

Duffie, who founded both ISDN innovator Origin Technologies Inc. and ATM/Ethernet switching startup Whitetree Inc. before forming Turnstone this year, has been interested in the practical issues of bandwidth availability throughout his career. It was that interest that had driven him at Whitetree to develop solutions for the practical quality-of-service problems that plagued the ATM and IP environments.

Duffie also was central to the development of some technologies-notably, the Integrated Services Digital Network and Switched Multimegabit Data Service-that didn't live up to their market potential. The market failure of ISDN and SMDS, however, are largely attributable to missteps by the phone companies.

Consequently, Duffie treats digital-subscriber-line provisioning very seriously. The incumbent carriers, he said, simply must get DSL right the first time if they are to retain any hope of attracting broadband customers.

Despite some false starts in Duffie's career, few of his chosen paths have led to nowhere. SMDS may have failed to take off, for example, but Duffie's work on the service's Distributed Queue Dual Bus architecture proved applicable to ATM. Similarly, the Whitetree concepts for providing universal services per port to enterprise customers could be applied to access customers at the edge of the WAN, providing a continuity of sorts between Duffie's Whitetree and Turnstone stints.

"If I pride myself on anything, it's not necessarily originality, but letting concepts and ideas from one environment be applied in a totally unrelated environment," Duffie said.

Duffie got his first taste of the real world-and liked it-via his summer stint at Bell Canada after graduating from McGill University, which had granted him a degree in engineering physics. He consequently looked for a permanent job that mixed practical design and development engineering with a vision for improving access for businesses and consu- mers. His first real job was with a Bell Northern Research group developing line cards for the Nortel DMS-100 switch; there, he learned how to use programmable digital signal processors to provide universal services to a broad range of customers.

Duffie later moved from BNR directly to Nortel in the late 1980s, and he still counts the Canadian manufacturing giant among his favorite employers.

"I could have stayed at Nortel a long time, but I was hot to trot to stretch my entrepreneurial legs in 1988, which is when I formed Origin," he said.

That company started as almost a network-service provider, offering a network-based backup service as a transitional vehicle to developing a remote-access concentrator based on ISDN. But carriers throughout North America proved unable to deploy ISDN successfully among sufficient numbers of customers. Duffie had no interest in continuing Origin as a consulting service, so he reached an agreement with the board of directors in 1990 to shut down the company.

From there, he went to NetExpress to work with Larry Roberts, a longtime ATM innovator who was the force behind the cells-in-frames ATM concept and many QoS ideas. The two executives made inroads into DSC Communications Inc. and ended up with a $14 million project to develop high-end multiservice ATM switches for that company.

The NetExpress and DSC work gave Duffie several ideas for optimizing enterprise data traffic by using both Ethernet and ATM. He went to see several venture capitalists in mid-1993. After working with such leading venture capitalists as Geoffrey Yang for six months, he had sketched out the concept for Whitetree.

The venture executives also introduced Duffie to Maureen Lawrence, who had managed the Adaptive Corp. ATM spinoff of Network Equipment Technologies Inc. Lawrence would become Whitetree's chief executive; Duffie served as chief technology officer.

The architecture for Whitetree's hybrid frame/cell switch was based on several premises for introducing ATM to corporate America. The first determination was that end-to-end ATM strategies were unlikely to get far. Corporations needed to have transition plans for LANs that were already in place, and they needed a reasonable broadband technology upgrade that could meet cost targets. There also needed to be a well-defined benchmark for performance and price at the edge of the network, the goal being to realize a hybrid switch port that was lower in cost than a switched Ethernet port.

"That led many of us in the ATM Forum down the primrose path of 25-Mbit ATM, and ATM25 was an absolute design nightmare," Duffie said. "I remember being in the standards meetings, hearing the chip vendors talk about 51-Mbit interfaces and saying that we needed to rely on a technology that was already in production, borrowing aspects from token ring. In retrospect, if I had really been awake at that point, I would have gone a step further and insisted on keeping the Ethernet physical layer and running ATM on top of it."

Duffie's team designed seven ASICs to do packet forwarding, cell switching and cell segmentation and reassembly. Whitetree did not do shabbily by anyone's account; its enterprise switch was recognized as a great Ethernet switch in its own right, and the company did $12 million worth of business in 1995.

Meanwhile, Duffie had been leading a group that was developing a next-generation ATM chip set with advanced QoS and queuing features. The company reached marketing pacts with General DataComm and Efficient Networks Inc. in late 1996, and it signed a broad chip-design agreement with Texas Instruments Inc.

But the bulk of Whitetree's 1995 sales had been to ATM enthusiasts-and by mid-1996, executives at Whitetree had reached a couple of core conclusions. First, ATM was not going to make it in the enterprise as a native switching system. Thus, the company needed either to reorient its business toward carriers, for which it did not have the channels or expertise, or to move to higher levels in the enterprise backbone.

To implement the second option, Whitetree had to be acquired. And since Ascend Communications Inc., before its acquisition of Cascade Communications Inc., had been looking for a good ATM architecture, Whitetree executives arranged the sale of the company to Ascend in April 1997.

In July of that year, Duffie decided that he wanted to take some time off and think about what he wanted to pursue for his next startup. He got married in August 1997 and then spent the fall doing part-time consulting and pondering his options.

"I wasn't even completely convinced I wanted to do another company," Duffie said, "but once you start a few, you get it in your blood. I had been looking at a few possibilities in DSL technology. Tom LeFleur, the man who started Rhythms [Networks, an Internet-service provider], talked to me about the limiting factors in DSL deployment. He said everything boiled down to truck rolls: How could carriers reduce the number of truck rolls by service personnel?"

At the end of 1997, Duffie took an extended vacation in Canada and spent several hours pondering the architecture of a typical central of-fice, identifying every entry point at which a copper pair is accessed and identified in the central office. The lack of visibility of a DSL-enabled copper pair shocked Duffie and several potential partners.

Duffie knew Eric Andrews, former vice president of marketing at Newbridge Networks Inc.'s Vivid division, from their joint participation in ATM Forum working groups. Indirectly, he also knew Andrews' boss, Vivid general manager Rick Tinsley.

"When Rick and I first started talking in the fall of 1997, we wanted to do gigabit frame relay, a little like packet-over-Sonet but not operating above Layer 3," Duffie said.

But as they observed the number of giga-router vendor start-ups doing packet over Sonet, Duffie and Tinsley worried about the business fundamentals, since too many WAN startups seemed bent on being acquired. Getting bought, Duffie said, is not a driving factor for him: "What turns me on is creating a new culture and attracting people to work for a fun and interesting place."

After Duffie discussed DSL possibilities with several carriers at year's end, he came back to Tinsley and Andrews with a far simpler concept: no ASICs, no high-layer switching, just a box that provisioned and allocated physical copper channels to DSL services.

Turnstone Systems incorporated on Jan. 10, 1998, and the three founders pledged to have a chassis-based electromechanical physical provisioning system for DSL ready nine months later. Duffie said he had to adopt a constant mantra to keep the Copper CrossConnect system simple and elegant. The result was the achievement of a testable system by October.

Keep it simple. Thoreau would have appreciated that.

Return to: TimesPeople 98