The early years of Tandem

Treybig got the idea behind Tandem while still working at HP, where he became a salesman for HP's first 2116 computer systems after joining the company right out of Stanford. His boss was marketing chief Tom Perkins, who left to form pioneering Silicon Valley venture capital firm Kleiner & Perkins, now Kleiner Perkins Caufield & Byers. Perkins is also Tandem's former chairman and was a Compaq board member.

Treybig followed Perkins, and became Kleiner Perkins' fifth employee, spending two years there working on the Tandem idea. At HP, Treybig saw customers lash two HP servers together so one would take over if the primary system failed. These customers would replace the standard operating system with custom-built software, causing all sorts of support problems. Treybig thought such a setup could be created with standard computer hardware and software, replacing these exotic one-off systems.

"You could build an operating system that wouldn't fail and would be standard software," Treybig said. But to make the idea feasible, the systems had to be designed with the goliath input-output capabilities of mainframe computers, he said.

Treybig launched Tandem in 1974, with half the first 18 employees coming from HP. He also hired three gurus to design the system. Mike Green had also been at HP, where he had created its first time-sharing computer, a system that could work on several jobs simultaneously by quickly switching between different tasks. Jim Katzman, too, had been at HP but had become head of mainframe maker Amdahl's input-output work. And Davie Mackie had been working on custom systems for discount retailers in Europe.

"The three of them put together the architecture of the Tandem. There's still nothing like it," Treybig said.

Indeed, Illuminata's Eunice said, Tandem succeeded at an extremely difficult task, one nobody has yet been able to do as well. Tandem's technology can simultaneously distribute a gigantic database across numerous independent processors while supporting high volumes of transactions.

"Tandem's the only one who whacked at vastly parallel databases and was able to do online transaction processing at the same time," Eunice said.

But Tandem couldn't sustain its early growth. By the mid-1990s, revenue had flattened out at about $500 million a quarter, and Tandem had undergone rounds of layoffs, cutting hundreds from a staff that had grown to 9,500.

"In the early days," consultant Bonhiver said, "Tandem had much better growth because there were so many different kinds of (computer systems) out there, it didn't stand out as this weird, niche platform. Now everything is all Sun or (Windows) NT and some mainframe."

Hewlett-Packard never offered to buy Tandem, but it's clear HP was always in the back of Treybig's mind. In fact, recurring visions of HP would come to Treybig in his sleep.

"I used to dream that we failed, and (HP) put me in a cage in the front entryway," Treybig said.

"It was my nightmare."

Treybig initially proposed a buyout to Compaq, thinking Tandem couldn't survive on its own and that Compaq's thrust toward services and direct sales matched Tandem well. But Compaq wasn't interested at the time, and Treybig left Tandem in 1996, the year before Compaq changed its mind. Treybig now is a partner at Austin Ventures in Texas.