Over the weekend the New York Times ran a piece about the upcoming award of the NSF petascale contract. The piece is based on
…documents that were accidentally placed on a federal government Web site for a short time last week.
and alleges that the documents show that IBM is going to build the Track 1 machine and that the machine will be placed at the NCSA.
The tone of the article is that there’s something afoul in the system awarding the machine
The award has been eagerly pursued by a number of supercomputer centers and state governments. Word of the decision to award the contract to I.B.M. to build a production version of a computer that is now intended for the Pentagon’s Defense Advanced Research Projects Agency has created widespread concern in the past week among some computer scientists involved in designing and building the nation’s high-performance computers.
I’m not sure why we’re conflating the DARPA HPCS program and the NSF procurement, or why there is any conflict there. I’ve read a few other blog postings on the internet that speculate that NCSA proposed their DARPA machine for the Track 1 build. If true this is risky, which is perhaps the point the Times was (obliquely) making. Also, HPCS machines obviously aren’t “intended” for DARPA, but that’s semantics.
The concerns seem to be about political influence in the process
Several government supercomputing scientists said they were concerned that the decision might raise questions about impartiality and political influence.
I’ll let you draw your own conclusions about whether there’s anything untoward going on. I personally think there’s way too little evidence to be speculating about anything.
The article also throws in some concerns about the Track 2 award.
There were also concerns expressed by researchers about a second award discussed in the documents. The N.S.F. would install a Cray supercomputer at the Oak Ridge National Laboratory at the University of Tennessee. Several researchers said they were concerned that although the award was to the university, its operation would be carried out by an Energy Department laboratory.
A researcher for the Pittsburgh Supercomputer Center who identified himself as “grouchyoldcoot” in a Web blog posting, last week wrote: “Having the N.S.F. buy a machine and hand it to the D.O.E. is, well, very unexpected — the D.O.E. has plenty of machines of its own, and the N.S.F. wants to get its work done, not the D.O.E.’s.” The posting was removed on Thursday, but not before it was circulated among supercomputer researchers.
Of course I haven’t read the details of the award so I don’t know what kind of relationship is intended. But it seems to me that it would make a very strong proposal from the evaluator’s point of view for a very large machine to be placed in a university — and used for NSF research — but run by a group of seasoned professionals who run very large machines for a living. It’s totally conceivable that the ORNL guys are effectively acting like subcontractors running the machine for UT and the NSF.
[…] NSF Petascale story in the NYT […]