Stanford Report, October 18, 2000
BY DAWN LEVY
John Hennessy's graduate students called him "The Big Man."
"That's capital T, capital B, capital M," says Mark Heinrich, Hennessy's student from 1991 to 1998, now an assistant professor in Cornell's School of Electrical and Computer Engineering.
"Hennessy is renowned and admired for his ability to see the Big Picture, and provide direction and advice so clear that problems you'd been grappling with for weeks disappeared," Heinrich says. "We would often wait outside his office for a legendary 'walk with me' meeting. 'John, do you have time for a few questions?' 'Walk with me.' While walking between appointments, he was able to steer you onto a corrective course of action in minutes. You often found yourself on some part of the campus where you didn't want to be, but that was a small price to pay."
Hennessy has scaled mountains as a researcher, teacher, entrepreneur, author and administrator, and appreciates both the devilish details upon which a system is built and the guru's view at the top. As a pioneer in computer architecture, he has built a career optimizing the flow of information, simplifying complicated tasks and minimizing failure -- great training for his new role as Stanford's 10th president.
"Some people seem to succeed in life by dark strategy and devious planning, and other people seem to just have a better idea and can put it simply so everyone else can't help but go along," says computer science Professor John Mitchell. "I've seen John follow the latter strategy flawlessly. If you think nice guys finish last, he's living proof to the contrary."
Hennessy joined the Stanford faculty in 1977 as an assistant professor. At the time, researchers in the field of very large scale integration, including Hennessy, were trying to place as many transistors as possible on a chip. Out of this work came reduced instruction set computer (RISC) architecture, which, compared to complex instruction set computer (CISC) architecture, uses simpler but faster instructions to execute the same task. Researchers at IBM, the University of California-Berkeley and Stanford produced the first RISC microprocessors.
Hennessy led RISC research here. His work focused on compilers -- software that translates programming languages like C into instructions that a computer can understand. With a handful of graduate students, he began a project in 1981 called MIPS (for millions of instructions per second) that set out to simplify computing with RISC architecture and ended up revolutionizing an industry that had relied on lower-performance CISC architecture.
Hennessy became an associate professor in 1983 and full professor in 1986. While on a sabbatical leave from 1984 to 1985, he cofounded MIPS Computer Systems (now MIPS Technologies). The company furnished chips to Silicon Graphics, one of three billion-dollar companies built by Jim Clark, former Stanford professor and Hennessy colleague. In 1992, Silicon Graphics bought MIPS for $333 million. In 1999, Clark donated $150 million to Stanford for interdisciplinary biomedical research. Hennessy orchestrated the donation -- Stanford's largest since its founding grant.
Hennessy's Stanford career demonstrates "an organizational form of Moore's Law, doubling his influence and stature within the university every 18 months," Mitchell says. (Named for Intel cofounder Gordon Moore, Moore's Law predicts that the number of transistors that can be packed on a computer chip -- an indicator of performance -- will double every 18 to 24 months.) Hennessy served as chair of the Computer Science Department from 1994 to 1996, dean of the School of Engineering from 1996 to 1999 and provost from June 1999 to September 2000. He has won numerous awards and is a fellow of the Institute of Electrical and Electronics Engineers and of the American Academy of Arts and Sciences and a member of the National Academy of Engineering.
He also is author or coauthor of more than 100 technical publications, including two popular textbooks he wrote with David Patterson, the Pardee Professor of Computer Science at the University of California-Berkeley: Computer Organization and Design: The Hardware/Software Interface and Computer Architecture: A Quantitative Approach. The latter, considered the classic introduction to computer architecture, begins with the Monty Python quote: "And now for something completely different . . . "
Quick and innovative integration
"John has always grounded his approach to computer architecture in engineering reality," says Chris Rowen, a former doctoral student of Hennessy's and president and CEO of Santa Clara-based microprocessor firm Tensilica Inc. "He brought together countless analytical tools, especially for modeling and simulation, to allow architects to make informed and dispassionate tradeoffs among approaches, both on a grand scale and in the minutiae of the design details."
The key to Hennessy's quick analysis of complex situations is his ability to listen and summarize, Rowen says: "He can follow a complex, sometimes even muddled, explanation of some analysis, extract the important facts, disambiguate the problem data and restate it all in a form that sheds useful light. This makes the whole team more effective in getting on to the next task."
Hennessy's experience with detailed processor analysis fine-tuned his intuition, says Rowen: "He could trot out both the relevant facts and the dominant technical trends to help drive a decision to conclusion. This remarkable 'back-of-the-envelope' analysis was not necessarily infallible, but so guided the design exploration that much more time could be spent on the most productive design ideas."
That intuition helps Hennessy make "incredibly bold predictions," Berkeley's Patterson says: "In 1986 or so he said that any company that didn't have a RISC machine for sale would be out of business in a few years. Within his time limit virtually every company offered a RISC product."
Teamwork and play
"John really built the computer systems group at Stanford," says Mark Horowitz, Yahoo! Founders Professor in the School of Engineering and professor of electrical engineering. "He hired us, found funding, pushed us in some interesting directions if needed and then stepped back so we got the credit -- sometimes a little more than we deserved."
Computer science Professor Monica Lam notes that she "appreciated him even more after he left [for the dean's office] as I realized how difficult it is to make a major research project like DASH [a large multiprocessor project] happen. It is not just about bringing in lots of money and having lots of smart people around. It takes research leadership and vision."
And hard work. But Hennessy also plays hard. Lam recalls arriving at a week-long conference midweek, only to find the students "extremely exhausted. They showed up late to the technical programs and were falling asleep in the conference." It turned out that these 20-something-year-olds were tired out from playing hearts with Hennessy late into the night. "In the meantime, John was just doing his usual things -- got up at the crack of dawn, and probably finished writing a chapter or two of his textbook -- before giving his keynote to kick off the technical program for the day," Lam says.
The vision: Ubiquitous access to information
Despite the demands of his position while provost, at this spring's Stanford Computer Forum, Hennessy delivered a keynote address outlining challenges to the industry. The good news? Computer scientists have convinced the world that it needs the Internet. The bad news? Now they must meet the demands created by their successes and ensure access to information anytime, anywhere. While it might be OK for web-savvy users to suffer frustration if they go to a web page and see a blank screen, the same poor performance won't be tolerated by a public used to getting Fox every time they hit Channel 2 on their TV remotes.
Seamless access to information means engineers will need to rethink how to make information access available, maintainable and scalable via wireline and wireless networks. But as the complexity of a task increases, system components are at greater risk of failure.
Like computers, universities including Stanford are increasingly engaged in complicated tasks -- extending education through distance learning, finding the functions of genes, bringing together researchers in diverse disciplines to solve new problems, finding affordable housing. Universities will make progress, gradually. Stanford may be in especially good hands with an engineer leader versed in minimizing failure and squeezing the best performance from a system.
While Hennessy's address to the Stanford Computer Forum dealt
with computers, not universities, computational and educational
systems have some similar responsibilities: efficient transfer of
information and execution of tasks. His comments about building
computers that serve their users well also may apply to his
presidency: "Somehow we have to live with these unreliable,
individual components and yet make a system that works for us."