SAN JOSE, Calif. Veteran supercomputer researcher Dan Reed will join Microsoft Research as its new director of scalable and multicore computing. Reed brings to Microsoft a long background in parallel computing and activism in technology policy matters, as the group gears up to tackle problems that have puzzled scientists for twenty years.
"There is a sea change in computing coming at the intersection of multicore and large data centers, and working on this is one of the most exciting things I can imagine doing," said Reed who most recently served as director of the Renaissance Computing Institute at the University of North Carolina.
"I've been working in parallel computing my entire professional life and the move to multicore has pushed this problem to the forefront of computer science," he added.
There's no single path to the parallel programming models needed to support tomorrow's multicore processors, said Reed. "It will take a variety of efforts in areas such as functional languages, transactional memory, extensions of existing languages and new higher level tool kits," he said.
Reed will collaborate with Burton Smith, another parallel computing guru who joined Microsoft in 2005 to help spearhead work on multicore issues. In an interview earlier this year, Smith discussed work on at least two functional languages at Microsoft.
"My primary mission is to look at the large data center problem with a secondary charge to look at many-core architectures. Burton Smith's has the same priorities in reverse order, so we will be working together on both issues," Reed said.
Large data centers face spiraling power and reliability requirements as they try to keep up with growing application loads driven by Internet-era applications such as Web services and social networking. "I want to take a green field approach to these issues," said Reed.
Reed will retain his seat on the U.S. President's Council of Advisors on Science and Technology. He will also continue as current chair of the Computing Research Association, a group of top academic computing departments and industrial research labs in North America.
"In these times of tight federal funding for R&D, researchers tend to move toward more short-term incremental projects," Reed noted. "One pressing need is to get more federal funding for long-term, large-scale and high-risk projects. The government also needs to do a better job of strategic planning across its agencies," Reed said.
For example, the industry has pretty thoroughly mined parallel computing concepts from academia. "We need to refresh that well," said Reed.
Reed served as director of the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign for four years. He was also chief architect of the National Science Foundation's TeraGrid, a national distributed computing system for researchers.
The move to Microsoft Research marks Reed's first job in the commercial sector. "IBM Research has a broader purvey because their work includes the basic physics of devices, but in the last decade Microsoft Research has emerged as the premier computing research organization in the world by any metric," Reed said.