March 18, 2008
Intel, Microsoft Give $20-Million to Universities to Push Computing Limits
Parallel awards to two research universities will help push parallel computing to the next level, executives at the Microsoft Corporation and the Intel Corporation said today. The University of Illinois at Urbana-Champaign and the University of California at Berkeley will share $20-million in corporate dollars over the next five years for a project that researchers say could produce pocket-sized intelligent devices that recognize people and tell you all about your past interactions with them.
That vision of a personal digital assistant with a visual memory “and that can whisper in your ear” was conjured up by David Patterson, a Berkeley professor of computer science. But it depends on parallel computing, or multiple computer processors working together on a single task divided up into many parts. Mr. Patterson will direct the new Universal Parallel Computing Research Center at Berkeley; the twin center at Illinois will be led by Marc Snir, a professor of computer science there.
The Berkeley group will comprise 14 faculty members and 50 graduate students and post-docs. The center at Illinois will have 20 faculty members and 26 graduate students. Both centers will try to develop software to coordinate these multiple processors, or multiple processing “threads” within a single unit.
That software, executives of the two corporations said, will be available to the companies for non-exclusive license if they want it. Tony Hey, corporate vice president of external research at Microsoft Research, said that would be a rare option. Most of the software rights will remain with the universities, who intend to distribute it with open-source licenses.
The stable source of money is welcome in a strapped financial environment, researchers said. “These days it’s really wonderful to have five years of funding,” Mr. Patterson said. —Josh Fischman
Posted on Tuesday March 18, 2008 | Permalink |Comments
Commenting is closed for this article.
Previous: E-Voting Vendor Threatens Princeton Computer Scientists With Legal Action
Next: Tech Therapy: 'If We Don't Pay Attention to IT, We're Dead in the Water.'
Twenty-million dollars spread over five years for 34 faculty members and 76 grad students at two universities? That will look nice on a couple of resumes, but, with high computer-science salaries and substantial indirect charges, it’s hardly the foundation for a product breakthrough. In addition, all this distributed work will convert much creative energy into coordination energy. Or maybe the PR department got out of hand, and the idea is simply to produce a large number of loosely related small efforts with numerous papers graciously crediting Microsoft and Intel.
If so, it’s industrial feel-good money that could be much-better spent — speaking as a shareholder — on a smaller, well-focused research group consisting of three or four faculty members, a handful of grad students and post-docs, and many-more full-time M.S.- and Ph.D.-level research associates (or Microsoft and Intel employees on assignment to Illinois and Berkeley). If you intend to “Accelerate Benefits to Consumers, Businesses,” as the Microsoft press-release promises, give the money to a highly talented and focused group of academics and research professionals.
I get nervous when companies claim they want to help advance a scientific discipline, but if that is the goal, then a pocket-sized intelligent device “that recognizes people” and tells you “all about your past interactions with them” is the wrong project, given the important problems we face in computing. Coupling massive parallelism with as-yet unknown concepts to extract information — not just data — from the incredible repository called the World Wide Web or finding a way to bring engineering sanity to the creation of large, complex software systems are two of the numerous reasonable projects that would both advance science and engineering and lead to better high-tech products. I speak, now, from my experience as a faculty member who has run $20 million software-research projects and knows where the money went.
— S. Britchky Mar 18, 06:11 PM #
Berkeley knows how to run such large projects successfully. Past projects that led to multi-billion dollar industries include: more efficient microprocessors (Reduced Instruction Set Computers or RISC), reliable storage systems (Redundant Arrays of Inexpensive Disks or RAID), and clusters of computers that are the foundation of Internet service companies (Network of Workstations or NOW). That research investment led to better computing systems that led to more jobs and US exports. In fact, an NAS study in 2003 (“Innovation in Information Technology”) identified 17 multibillion dollar industries in the IT field that came about from university research, and Berkeley was associated with 7 of them, more than any other university.
Remember that $2M per year is matched by state funding, which allows us to create a single space covering many disciplines and graduate students working on this critical problem. Given Berkeley’s track record and the people involved, many reasonable people think its a sound investment for Intel and Microsoft stock holders. Some publications even wonder why they don’t do even more centers, given the importance of the problem to future of the companies.
If no one solves this problem, programs will not run faster on newer computers, which is something we have enjoyed for more than 30 years. What portion of the improvement in productivity in the US economy relies on the improvement in cost-performance of computing? What will be the impact on the IT companies and the US economy if computing performance stops improving?
— David Patterson, UC Berkeley Mar 18, 09:09 PM #
At Illinois, we already have many ongoing small-group projects on solving this problem. These projects will continue to create breakthrough technologies. But we believe that parallel computing is at a juncture today where the solutions need to be driven by the whole system stack. This requires bringing together a diverse group of researchers in applications, languages, compilers, software engineering, programming tools, formal methods, runtime environments, and hardware. This is the vision behind Intel and Microsoft coming together and is the vision driving the Illinois center. More than the funding, a center that has the ear of key industry partners provides a very strong incentive for such diverse faculty to come together and rework their agendas in a much larger context. This synergy is a key benefit of such an effort. Illinois is particularly well suited because of the breadth and depth of our existing research in all of these areas, and our established record of innovation in parallel computing (the NAS report that Dave sites credits Illinois with starting the parallel computing industry).
As for applications, information extraction is a large part of the application effort that drives our work and software and system robustness is a key goal.
Sarita Adve
Director of Research
UPCRC/Illinois
— Sarita Adve Mar 21, 12:50 PM #