CRA joined 31 other scientific societies and universities yesterday in showing off the results of NSF-sponsored research at the 10th annual Coalition for National Science Funding Science Exhibition and Reception on Capitol Hill. CRA was ably represented at the event by DK Panda and his students (Jiuxing Liu, Pavan Balaji, Ranjit Noronha, and Sayantan Sur) from The Ohio State University, who presented work on software that allows high performance, scalable communication using the InfiniBand networking technology.
The Exhibition was a great opportunity for making the general case for federal support of basic research, especially at NSF. This year’s event was widely-attended. Many key congressional staffers, influential Members of Congress, and important members of the Administration and NSF took time out of their schedules to see the exhibits. Here’s the proof! (click for larger images)
|The CRA booth. Underneath the table were four PCs clustered together with InfiniBand. Two monitors show the results of some benchmarking apps comparing InfiniBand to Gigabit connections.|
|NSF Director Arden Bement (left) listens to Professor Panda describe his research.|
|Ohio congressman Dave Hobson (R-OH), a very influential member of the House Appropriations Committee. Hobson has the distinction of being the only member to serve on the appropriations subcommittees for Defense, VA-HUD-Independent Agencies (home of NSF funding), and Energy and Water (which he chairs).|
|Hobson takes some time to speak with Panda’s students.|
|Director of the White House Office of Science and Technology John Marburger stops by the CRA booth.|
|Marburger also took time to speak with the students and ask some additional questions about the research.|
Thanks again to Professor Panda and his students!
The Senate Committee on Energy and Natural Resources held a hearing Wednesday on a bill to authorize the building of a “Leadership Class” supercomputer at DOE. S. 2176, the High End Computing Revitalization Act of 2004, introduced by Senators Lamar Alexander (R-TN) and Jeff Bingaman (D-NM), would authorize $250 million worth of HEC R&D at the Department of Energy through FY 2009 ($40M in FY 05, building to $60 M in FY 09); it would authorize $500 million through FY 09 ($100 M a year from FY05-FY09) to construct a new supercomputer with “100 times the capability” of the fastest computer in existence at enactment; and it would authorize $50 million through FY09 ($10M a year) for the creation of a High-end Software Development Center. The money authorized would be “new” money, so appropriators would have to come up with additional money to fund it.
The bill is very similar to a House Science Committee bill introduced by Reps. Judy Biggert (R-IL) and Lincoln Davis (D-TN). Both are loosely based on the recommendations of the HEC Revitalization Task Force Report. The House bill contains fewer authorizations — no specific authorization for a software development center, for example — and more modest authorization levels: only authorizing $50 million in FY 05 for HEC R&D, building to $60 million in FY 07. Administration sources tell us that the President would likely sign the House version if presented to him, but would have serious problems with the more ambitious funding requests in the Senate version. The House version of the bill will likely head to the floor of the House in July for approval (along with the Science Committee’s broader HPC Reauthorization bill). The future of the Senate bill is less certain.
However, the Senate hearing was a good opportunity to get comment on the bill on the record. Testifying before Senators Alexander and Bingaman were James Decker, Principal Deputy Director of DOE’s Office of Science, (filling in for Office of Science Director Ray Orbach); Jeff Wadsworth, Director of Oak Ridge National Lab; David Turek, Vice President, Deep Computing at IBM; Dan Reed, CRA Board Member and Director of Renaissance Computing Institute at UNC-Chapel Hill; Vincent Scarafino, Manager of Numerically Intensive Computing at Ford Motor; and Dimitri Kuznezov, Director of Advanced Simulation and Computing at DOE NNSA.
As tends to happen lately when Members of Congress discuss high-end computing, much of the focus of the hearing centered around what could be done to recapture the “supercomputing lead” from the Japanese and their Earth Systems Simulator machine. In his opening remarks, Alexander noted that the ESS seemed to indicate that “Japan is king” of high-speed computing, a fact that had lead both Alexander and Bingaman to travel to Japan to be briefed on the new machine. Alexander said that he learned in the briefing that many US researchers and companies had requested time on the Japanese machine, marking the first time he could recall that US researchers were looking overseas for computational resources — a worrying precedent. His bill, he said, is focused on recapturing that lead.
Bingaman echoed many of Alexander’s comments, noting that he’d always considered supercomputing to be “one of the long poles in the tent for US leadership in science and technology.” He noted later during the question and answer period that high-end computing was “one area I’d prefer us not to have to outsource.”
Of particular concern to both Senators was the availability of both machines to US researchers not in DOE. “We’ve got a secret weapon [in the US] called our research universities,” Alexander said, noting that providing access to those universities should be an important piece of the supercomputing effort at DOE. Decker assured the Senators that time on any new “Leadership Class” machine would be merit-based, peer-reviewed and competitively awarded. Neither he nor ORNL’s Wadsworth mentioned what percentage of cycles would be available to university or private sector researchers.
The hearing was basically non-controversial, with both Senators promising to try and move forward with their bill. As that process moves forward, check this space for details.