On Tuesday the House Committee on Science and Technology’s Subcommittee on Research and Science Education held a hearing in the Rayburn House Office Building to gather information and perspectives on the research infrastructure needs of universities and colleges as part of the push to new the America COMPETES Act.
From the committee press release
“Our focus on this legislation is a direct acknowledgement of the fact that America’s science and technology enterprise underpins the long-term economic competitiveness of our country. The partnership between the federal government and our nation’s colleges and universities has been highly successful and has led to a great number of societal and economic benefits,” stated Subcommittee Chairman Daniel Lipinski (D-IL).
…The physical infrastructure for research includes not only bricks-and-mortar buildings, but also research instrumentation and a robust cyberinfrastructure. Cyberinfrastructure, which consists of computing systems, data storage systems, data repositories, and advanced instruments, has become increasingly important to all science and engineering disciplines. The Office of Cyberinfrastructure at NSF requested a budget of $228 million in FY 2011, a 6.4 percent increase from FY 2010.
Witnesses at the hearing included Dr. Leslie Tolbert (University of Arizona), Mr. Albert Horvath (Pennsylvania State University), Dr. John R. Raymond (Medical University of South Carolina), Dr. Thom Dunning (National Center for Supercomputing Applications). You can see all of their prepared statements, as well as Chairman Lipinski’s opening remarks, at the hearing web page.
Dunning focused much of his comments on the high performance computing part of the national research infrastructure, as might be expected. His written statement is (long but) interesting. I was particularly interested in his views on the state and impact of the NSF’s cyberinfrastructure efforts (the NSF falls under the House Committee on Science and Technology)
NSF has been successful in deploying new computing systems that are delivering extraordinary value for the U.S. research community — the first system delivered to TACC exceeded the total computing capacity of NSF’s TeraGrid by a factor of more than 5. However, the focus of these acquisitions was on the delivery of raw computing cycles and the funding available to provide support for the users of these new high performance computer systems was limited. This is unfortunate because this approach favors those scientists and engineers who are already using supercomputers and need little assistance, while our experience at NCSA and that at many other centers indicates there is a growing need for high performance computing resources in almost all fields of science and engineering.
Without adequate user support, it will be difficult for these new researchers to make effective use of the available resources. High quality support staff is one of the most valuable resources in NSF’s supercomputing centers and a fully funded user support program is needed.
This goes back to what is becoming an regular criticism of the NSF’s computing efforts (see recent comments by Smarr and Karin): most of the funding is for the capital acquisition, with not nearly enough support for operations and sustainment. Dunning also adds to the chorus of complaints about the way the NSF runs the cyberinfrastructure program
It should also be noted that the prospect of continual competitions has a corrosive effect on the staff at the centers — it is not only difficult to hire quality staff with funding that only lasts for 4‐5 years, but enormous amounts of staff time have to be dedicated to preparing for the competitions, rather than assisting researchers. The advantages of competition must be carefully balanced against those of stability in NSF’s supercomputing centers program.
Dunning’s statement also discusses the need for an increased emphasis in software, and identifies “the dwindling number of supercomputer vendors in the U.S.” as an area of concern.
Of the remaining companies, only IBM and Cray are actively involved in research and development on supercomputing. Although you would have to talk with these companies to better understand the issues surrounding this situation, it is clear than the supercomputing industry in the U.S. is not as healthy as it was just a few years ago.
I’d have to give it more thought, but I think Dunning is significantly shortening the list of supercomputing vendors. First of all Convey has to be included, I think, and what about the Tier 1/2 cluster vendors that are out there serving a significant market? Dell? Penguin? What about the supercomputing infrastructure R&D that HP is heading up? What about Intel’s exascale efforts? And you can’t ignore NVIDIA.
I think what Dunning is really observing is that the US HPC industry is experiencing radical change — what my friend Joe Landman calls “creative destruction“, and I don’t think we can say at this point that the change won’t be for the better. I’ve not heard anyone make a case that the upheaval of the mid 1990s resulted in less effective computing.
I won’t argue that the Tier 1 majors of yesteryear are less healthy than they once were, but the vendors are in part to blame. Technology shifts aside, as a group the old line HPC vendors continue to chase marquee business at zero or low margin, a practice which depresses prices throughout the community for all vendors. One possible response to customers who want to continue growing the top of the Top500 list at roughly 5 times the growth of the average purchase price of its systems? Don’t bid.
As I finish up this post, I want to point out that I’ve highlighted a few sections of Dr. Dunning’s testimony to which I felt I had something to add, but I don’t want to suggest that I disagree with most of what he had to say. It’s far easier to edit than to create, and as they say everyone’s a critic. I appreciate the leadership and visibility Dunning and the rest of the witnesses are bringing to these issues.