In response, to the High-End Crusader article, Hard Questions While Waiting For The HPCS Downselect (http://www.hpcwire.com/hpc/645107.html), Raymond Turney writes:
Dear Editor,
The High End Crusader does not address the main issue involved in launching a major new high end computing architecture effort at this time.
We are apparently in an era of declining spending on research in the U.S.; particularly in C.S.; at the same time developing and effectively exploiting new HPC architectures will require a lot of resources {it usually takes at least 5 years to refine and figure out how to use a new technology – Globus and SRB date back to 1997 at the very latest, and after a lot of hype people are only now starting to figure out how to make Grid technologies work on a production basis, and figuring out the best way to use them}. Where will the resources to fight the battles involved in implementing heterogeneous architectures come from? As the author notes, the mission agencies have computational problems they need solved yesterday. As the author does not note, if you approach the mission agencies suggesting that they accept a budget cut to pay for development of machines which might be capable of solving them, you will likely meet with opposition.
In short, we are in an environment where coming up with the resources to push through a major HPC architecture development effort is going to involve goring a lot of oxen, and these oxen have owners. Advocates must show not merely that it is a good idea; but that it is a significantly better idea than alternative uses of the money. After all, money is a scarce resource; even more scarce than low latency bandwidth networks:-).
If he is intending to argue that Cray, because it does not have a large business involving commitment to traditional architectures, is likely to be more committed to new architectural development than Sun and IBM, he is probably correct.
I do agree with his call for disclosure of proprietary info for which companies are receiving government funding. We're talking about research here, where peer review is traditionally considered essential; not the code for MS Word.
Ray,