Still Boldly Going
For Rick Rashid, the voyage from Iowa to building the world's largest computer science research organization went through Carnegie Mellon
By Jason Togyer
(Dan Hart illustration)
It was a familiar sensation for graduate students working on the Mach kernel in the 1980s --- grinding out code under the fluorescent lights of Wean Hall, they would suddenly feel the presence of Rick Rashid behind them. Rashid, then a professor in the computer science department, hated meetings. Instead, he engaged in what his proteges jokingly remember as "managing by wandering the halls."
"Rick would tailor the frequency of his visits based on how he thought you were doing," Rich Draves (CS'94) says. "If you were seeing a lot of him, it meant either that he was excited, or worried. And that's still the case."
Draves is still working for Rashid, now senior vice president of research for Microsoft Corp. When Rashid left Carnegie Mellon in 1991 to set up the software giant's initial research effort, Draves was among his first 20 hires. Today, Microsoft Research includes 850 people --- including Turing Award winners Tony Hoare and Butler Lampson, and, until his disappearance and presumed death, Jim Gray --- and laboratories in Bangalore, Beijing, both Cambridges (Massachusetts and England), and Mountain View and San Francisco, Calif.
Under Rashid, Microsoft is investing heavily --- about $6 billion annually, according to estimates in trade publications --- into research into computer vision, e-Commerce applications and Web services, including search engine technology. Work done at Microsoft Research has led directly to creation of:
"Rick Rashid, to his credit, has been able to structure Microsoft Research in a way that's incredibly relevant to software research," says another of Rashid's former students, Jeff Eppinger (CS'88), one of the founders of Transarc Corp. and currently professor of the practice in Carnegie Mellon's Institute for Software Research.
- Microsoft Virtual Earth and TerraServer;
- Data mining and machine-learning tools on the company's SQL Servers;
- Compression and encoding algorithms of Windows Media Player; and
- Rendering technology and artificial intelligence inside Xbox.
"There's a saying in the industry, 'Don't bet against Steve Jobs,' and I'd say, 'Don't bet against Rick Rashid,'" Eppinger says.
Despite Microsoft Research's extraordinary growth, Draves, now area manager for systems and networking, says the organization retains a Carnegie Mellon flavor. Rashid still tells researchers about the "reasonable person principle" --- the SCS tradition favoring initiative over bureaucracy, as long as the results don't infringe on other people's rights. "The working atmosphere here feels very academic," Draves says.
That's no accident, says Rashid, who modeled Microsoft Research after Carnegie Mellon's computer science culture. His approach to building the organization has focused on finding researchers who can make an impact in their fields; collaborate with others from different backgrounds; and thrive when given freedom to pursue ideas.
"And I look at their curiosity --- do they pull in knowledge from other areas?" Rashid says. "Are they excited not just about the specific things they're doing, but is computer science as a discipline exciting to them? ... If you don't have that attitude that 'I'm going to keep moving and stake out new territory intellectually,' you're not going to have a long career."
Maybe it's no surprise that Rashid --- a proponent of boldly exploring new frontiers --- is a huge fan of the "Star Trek" franchise. He identifies with the thoughtful management style of Captain Jean-Luc Picard of "Star Trek: The Next Generation," who delegated authority rather than barking commands, and even has a Starfleet uniform like the one Patrick Stewart wore on the TV series. During Rashid's tenure at Carnegie Mellon, he started taking his researchers to the premiere of each new "Star Trek" movie --- at his own expense. (For the debut of the latest film, the tradition continued, though it required renting a theater for more than 500 people.)
And in the best "Star Trek" tradition, Rashid seeks out researchers who are invested in the organization, not just their personal goals. "It's a Carnegie Mellon way of doing things," he says. Eighteen years after leaving Pittsburgh, Rashid remains close to the university. His son, Daniel, is a doctoral candidate in the Language Technologies Institute, and the new Gates Center for Computer Science includes the 250-seat Rashid Auditorium, created with help from a gift to Carnegie Mellon from the family.
"I owe a lot of my personal success to the time I was at CMU, and I thought it was important that to give back to the university, and the students and faculty," Rashid says. "When I go back to Carnegie Mellon, it's the closest thing I have to going home."
Home originally was Fort Madison, Iowa, population 14,000, where his parents, Farris and Ramona Rashid, ran a wholesale grocery business. Located on the Mississippi River two hours south of the "Quad Cities," Fort Madison's biggest employers were the Iowa State Penitentiary and the Sheaffer pen company --- not exactly hotbeds for scientific research.
Nevertheless, Rashid's sister, Norma, a former television newscaster who now teaches journalism at the University of Cincinnati, says her older brother was always fascinated by science. As president of the science club at Fort Madison High School, Rick Rashid once locked himself in a laboratory so no one could interfere with an experiment, she says.
"Whatever he did, he was intense about it," Norma Rashid says. "When he was in high school, he put together an artificial heart for the science fair --- most kids grow bean plants, but that was the kind of thing he did." She remembers avidly watching television coverage of America's space race with her brother, and his interest in science fiction --- from writing his own stories to falling in love with "Star Trek."
"Rick was always interested in learning more about problems, and asking questions," Norma Rashid says, "and he's always loved science fiction --- so if you're going from fantasy to real science, the logical pathway is through computers."
Rashid didn't see his first computer (a Hewlett-Packard 2116) until his sophomore year at Stanford University, where he was a mathematics and comparative literature major. Fellow student Dan Ling --- later a Microsoft Research employee --- encouraged Rashid to take a basic course in Stanford's Algol-W programming language.
His first opportunity to program a computer was "a rush," Rashid says, and instead of pursuing a higher degree in math, he decided to go to the University of Rochester to earn his master's and doctorate in computer science. When his mother found out, she cried. "For someone from Iowa, I might as well have said 'I'm going to go get a graduate degree in underwater basket-weaving,'" Rashid says, adding his parents, both now deceased, were "never anything but supportive of me."
At Rochester, Rashid worked on research into the emerging field of personal computing over local-area networks, which attracted the attention of Carnegie Mellon's Raj Reddy. Reddy was spearheading what became the Scientific Personal Integrated Computing Environment project and in 1979 asked Rashid to join the faculty. One of SPICE's goals was a so-called "three-M" workstation --- a personal computer with a megabyte of memory, a megapixel display and a processor capable of handling a million instructions per second.
SPICE spawned the Accent operating kernel, whose novel use of memory mapping saved processor time by copying only the data needed for specific operations rather than complete files; and allowed processors to share information via secured ports. But it wasn't compatible with UNIX and was tightly woven around the Three Rivers PERQ workstation, limiting its usefulness. Rashid and others began work on the Mach kernel, which incorporated Accent's ideas but could run UNIX applications.
In a building full of computer scientists used to late hours, Rashid, lead developer on Mach, stood out in part because he arrived "bright and early" every day, remembers one of his graduate students, David Black (CS'88,'90). He also remembers Rashid's persistent optimism and encouragement to students. "He was an adviser in the best sense of the word," Black says.
"Tireless," adds Dave Hornig (CS'84), one of Rashid's advisees and now the vice president of product development at Pittsburgh-based Touchtown Inc. Rashid "clearly thought about computers from the moment he woke up in the morning until the time he went to bed," Hornig says.
Unlike other scientists consumed with finding elegant solutions, Rashid was most interested in effective ones. "I tended to have some instincts toward formalisms in theory, where he had the instinct to try something, see if it works, and if it doesn't, try something else," Hornig says.
Black says Rashid stressed the value of learning by doing. "One of the goals of the Mach project was to show that these concepts would work in practice, not just in principle," Black says. As a result, faculty and staff were soon running Mach-based operating systems on their own computers --- allowing the Mach developers to "eat their own dog food." They didn't always like the taste, especially when they found bugs, but that didn't discourage Rashid.
"I never saw him looking depressed --- maybe sometimes a little frustrated," Hornig says. "He had a can-do, positive attitude that rubbed off on a lot of people. In a place that was often rather hectic and sometimes negative, he could make you feel like it was all going somewhere."
Though compatible with UNIX, Mach wasn't hampered by some of the older operating system's restrictions. While UNIX allowed only a single thread of execution in a process, Mach supported multiple threads, sharing memory but running independently. That made it easy to expand Mach to run on multiple processors or across clusters of machines. Mach's modular design also kept the kernel size small while allowing to be ported to any number of platforms.
"Rick's work was really seminal," says Eppinger, because it successfully demonstrated ideas that became widely discussed among OS developers. And industry took notice --- Apple co-founder Steve Jobs visited Pittsburgh, left impressed, and recruited Rashid's students, including researcher Avie Tevanian (CS'85,'88) to his new company, NeXT, where the Mach kernel became the heart of the NeXTSTEP OS.
Microsoft's chief technology officer, Nathan Myhrvold, also came calling --- to recruit Rashid to set up a research unit, much to Rashid's surprise. "There were no software companies that had basic research groups in those days," he says. "Moreover, Microsoft had a relatively small number of products at that point. I had to be convinced that Microsoft would still exist in a few years, that it was willing to make the investment, and that we had enough 'intellectual runway' to get off the ground."
Before leaving Carnegie Mellon, Rashid set two key conditions. First, Microsoft's research unit had to advance the state of the art in any area it explored. Second, it had to turn that research into useful products. "It's not really that much different from Carnegie Mellon," Rashid says. "It's about developing ideas, and then getting those ideas to corporations, and the general public." Assured by Myhrvold that he'd have the freedom to pursue those goals, Rashid says, he made the leap.
Although industrial research organizations aren't unusual --- IBM established its first lab in 1945 --- Eppinger says Microsoft stands out as the only software company funding basic scientific inquiry. Microsoft scientists working under Rashid have an unusual amount of latitude to pursue research not directly tied to products.
Take computational biology: Microsoft researchers are experimenting with programming languages to alter living cells for use in medical therapies. Another effort is adapting algorithms used for digital compression and e-mail filters to model the ways that immune systems fight off intruders. Such research could eventually lead to a vaccine that adapts to the many mutations of the virus that causes AIDS. Other teams are pursuing research into fundamental theories of information retrieval, natural language processing, organizational behavior and human-computer interaction.
"The great contribution to mankind is when they let you go off and focus on something fundamental, but the great benefit for the company when you create this level of expertise is that the research group becomes a resource for the product centers of the company," Eppinger says.
Microsoft Research also has an unusually strong record --- at least among for-profit corporations --- of collaborating with outside scientists. Carnegie Mellon DNA is apparent in the collaboration process, which is governed by the RPP, Draves says. "In a lot of companies, when you have a paper that you want to publish, you have to go through the legal process and get various people to sign off on it," he says. "We don't have that process here. The onus is on the researcher to do the right thing. Go ahead and give talks, but be responsible about what you talk about." Microsoft Research investigators also publish frequently --- more than 5,100 peer-reviewed papers so far, including well over 10 percent of those accepted annually by SIGGRAPH since 2002.
After nearly four decades in computer science, Rashid says he remains excited about the frontiers --- especially cloud computing and what he calls "disembodied computation," where an end user's device delivers a service rather than processing a lot of data locally. Data-intensive super computing is another major focus for Microsoft Research, he says. "It's refined our notions of what our specifications for programs should be," Rashid says. "It's given us a better understanding of how we have to describe and talk about structuring data, so that we do a better job of sharing it and mapping it."
Whatever Microsoft Research pursues, Rashid says its main focus has to remain on people. "Some people think research is all about technology," he says. "It isn't. Research is all about people who have the ability to think and innovate. I think it's because we built that kind of organization with those kinds of values that we've been successful, both in the early days and in the long run."