Michael Stonebraker: The Database General

ichael Stonebraker sees himself as an evangelist of sorts for the high-tech world. As a founding father of the relational database, he's been driven by a belief that improvement is never out of reach.

"I guess I'm fundamentally what you would call a missionary, which [means] I think there's a better way to do things," Stonebraker says. "[For example], if you step up to 100,000 feet and look at information processing, we can do a lot better job than we're currently doing."

Stonebraker, 57, has been an architect of database technology and a catalyst in its evolution for the past 30 years. He has split his time between the academic and commercial realms, brainstorming his concepts in academia and then hatching start-ups to commercialize his ideas.

"I get bored selling conventional, mainstream ideas to mainstream people," Stonebraker says. "It's much more fun selling new ideas out on the bleeding edge. That's what really turns me on."

Friends and colleagues say Stonebraker's blend of intellectual curiosity and drive for progress that fueled his pioneering work in the database arena.

THE DATABASE GENERAL In pioneering the development of the relational database, Stonebraker drifted in and out of the academic and commercial realms, unearthing and cultivating ideas in academia and then applying them in the high-tech world through start-ups. He has embodied the high-tech sector's never-ending drive for new ideas and, in turn, progress.

Stonebraker graduated from Princeton in 1965 with a degree in electrical engineering. He says he had four options after finishing college: Vietnam, jail, Canada or graduate school. "That was a pretty easy choice to make," he says. He went to the University of Michigan, earning master's and doctoral degrees in computer science.

Stonebraker's journey in the database field began in 1971, when he joined the faculty at the University of California at Berkeley,just after IBM researcher Ted Codd wrote a now-famous paper on relational databases. The paper motivated Stonebraker to form a team, including students and colleague Eugene Wong, to work on developing relational databases, which would enable users to gather specific items of information based on the relationships between different fields of data.

Yet Stonebraker didn't pursue databases for any particular reason. "I just decided databases were a good place to go make a mark," he says. "It was kind of happenstance."

Stonebraker named the project Ingres, after a French impressionist painter. Ingres, however, was not the only relational database effort going on at the time. IBM was running a relational database project, known as System/R, which also sought to capitalize on Codd's thesis. One of the obstacles to implementing Codd's vision were the query languages he proposed, Stonebraker says.

"What you needed to do was propose query languages that real people could understand and figure out how to implement them efficiently," he says. "Ted Codd is a brilliant guy, but he's fundamentally a mathematician, and he was proposing predicate calculus-like query languages that were hard to understand. So you just had to build simpler ones. And then you had to figure out how to build an execution engine that ran fast."

That was the basic contribution of Ingres and System/R, which had a "synergistic relationship with one another," Stonebraker says.

In 1979, Stonebraker realized he needed to commercialize Ingres. Arizona State University had agreed to put nearly its entire student records application on Ingres but balked at continuing the effort because there was no organized support for the system's Unix operating system and Ingres database and the solution didn't support COBOL. So in 1980, Stonebraker founded Ingres Corp., which prospered selling relational databases to business customers.

During the same time period, another database entrepreneur emerged who would become a longtime competitor and nemesis to Stonebraker. In 1978, Larry Ellison formed Relational Software, which later became Oracle.

Stonebraker points out "an interesting and little-known fact": Oracle and Ingres were similarly sized in 1984. At the time, though, IBM decided to make DB2 its main database product for the MVS operating system. DB2 and Oracle supported SQL, but Ingres didn't. "So Larry Ellison had SQL, and Ingres didn't have SQL," Stonebraker says. "In the next two years, Oracle capitalized on the fact that they had SQL and Ingres didn't to leap vastly ahead in company size. So I think Larry Ellison was in the right place at the right time."

Stonebraker says he isn't fazed by Ellison's financial success. "Fundamentally, if Larry Ellison had had QUEL and Ingres had had SQL in 1984, then the ultimate destiny of Ingres and Oracle would have been reversed," he says. "Standards happen. So I think he got a really lucky bounce and took advantage of it."
Computer Associates International acquired Ingres in 1994, but well before that, Stonebraker had returned to academia and had spearheaded other advances in database technology.

At Berkeley, he worked to produce an "object relational" database system, which could handle more than numbers and character strings. The research project was known as Postgres.

Out of the research from the Postgres effort, Stonebraker founded Illustra Information Technologies in 1992. Informix acquired Illustra in 1996, and Stonebraker became Informix's CTO, a position he held until September 2000.

	ACHIEVEMENTS
	Today, technology areas that catch Stonebraker's eye include wireless and data integration on the Web. Started Ingres project in early 1970s at Berkeley to develop relational databases. Ingres Corp. formed in 1980. Another Berkeley project, Postgres, yielded object relational databases and spawned Illustra Information Technologies in 1992. Became Informix's CTO in 1996, holding that post until September 2000. Launched Cohera, a maker of federated databases, in 1999, based on a Berkeley research project, Miraposa.

If he wasn't occupied with developing database systems, Stonebraker says he might be interested in pursuing "sophisticated visualization" or working on "a combined applications and data design methodology." Data integration on the Web and wireless technology are areas that catch his eye, but "right now, I'm focused on seeing Cohera successful and making federated databases a reality," he says.
Teaching also remains a passion. Stonebraker says he's in the process of being appointed to the faculty at the Massachusetts Institute of Technology (MIT). He recently moved to New Hampshire from Silicon Valley, thus severing his ties with Berkeley.

"I really enjoy teaching," Stonebraker says. "The students at Berkeley,and hopefully at MIT,just challenge everything and really keep you on your toes and take your bad ideas and shred them. I think it's a great climate for exploring new ideas."

Jerry Held, a consultant at venture capital firm Kleiner, Perkins, Caufield & Byers, who worked with Stonebraker as one of the early Ingres architects, says Stonebraker knows how to fire up his students.

The teacher wouldn't hesitate to put students' ideas to the test, say Paula Hawthorne and Mike Olson, former students of Stonebraker. Hawthorne, now a consultant in Berkeley, says students made sure their ideas were on solid footing lest Stonebraker label them "stupid." Olson, who now directs marketing at Sleepy Cat Software in Berkeley, says students feared challenging the professor because they "were probably wrong," though Olson says he successfully challenged Stonebraker once.

Despite always yearning for the next project to advance the technology, Stonebraker says he feels fulfilled with the path he has chosen. "Computer science is among the most intellectually stimulating, fast-moving, complicated areas. I think it's ground zero right now. I think we will look back in 20 years to right now as the golden age of computer science because the Web is just completely changing everything. It's a really exciting time to be contributing to the field."