Microsoft is announcing today the appointment of Daniel A. Reed to the impressive title of Director of Scalable and Multicore Computing. It’s an interesting announcement for a number of reasons.
Reed comes from heading an organization called the Renaissance Computing Institute, a collaboration between the state of North Carolina and its three so-called Research Triangle schools—the University of North Carolina at Chapel Hill, Duke, and NC State—that, according to a press release, “focuses on finding solutions to complex, multidisciplinary problems and exploring the interactions of computing technology with the sciences, arts and humanities.”
Reed is also “a member of the U.S. President’s Council of Advisors on Science and Technology, and is the current chair of the Board of Directors of the Computing Research Association.” Microsoft says it wants him to continue to work on information technology research policy.
In a phone interview, Reed says he’s going to focus on the two big problems that his title names, the problems of large data centers and the software challenges of multicore computing, especially as the number of cores rises from 4 and 8 to 100 or more. “Clocks speeds aren’t going to rise significantly, so our advances will come from more cores.”
The issue has been handled so far largely through multithreading, but Reed says new methods are soon going to be needed, and “there isn’t a silver bullet. Like most complex problems, it won’t have a single solution.”
He divided the data center problem into that of reliability and scalability, on the one hand, and the environmental and other physical problems created by putting so many computers in one place. “It’s an important problem,” he says. “As we move to a service model of computing, the rate of growth of these data centers is phenomenal.” (In addition to Reed’s hiring, Microsoft announced this week a pair of $500 million investments in data centers, one in Illinois and the other in Ireland.)
He sees some of the answers as coming from the same forces that are creating large databases in the first place—sophisticated control systems, large numbers of sensors, and data analysis. “We’re using some of the results of sensors and computational fluid dynamics, for example, to analyze and model the air-flow in data centers.”
There are three things to notice about Reed’s appointment.
First, the position reports directly to Rick Rashid, the head of Microsoft Research. That says that Microsoft considers this area pretty important.
Second, that Reed is moving to Microsoft, instead of, say, Intel or IBM. Microsoft Research is, in his opinion, “the best computing research organization on the planet.” The research, he says, needs to be done in a company that itself has large data centers, and works with other companies that do as well, “and there aren’t that many of those.” Finally, he says, “these are in large part software issues, and while I have some background in architecture, I’m first and foremost a high-end software person.” He will be assembling and managing a team, but plans to “roll up my sleeves as well.”
Third, Reed is moving to the corporate world after 25 years in academia—20 years at the University of Illinois, where Mosaic, the first Web browser, was invented, then the last 5 in the Tarheel state. He says that the twin challenges of large data centers and multicores are creating the biggest tech revolution since the Web. “I watched the first revolution at Illinois… I resolved I wouldn’t watch the second one on the sidelines.”
