|
|
| News digest 25 June 2004 |
|
Advertisement
|
 |
| |
| Start |
|
|
| PrimeurLive! from ISC2004 in Heidelberg |
|
This year we again report Live! from Europe's main supercomputing event. See the issues for each day:
This year we start each issue with an experimental "blog" giving you some personal impressions. The next sections provide traditional reporting. The last section "Company news" provides mainly press release type of information from the companies. Hence, a few items are brought to you from three different view points.
|
|
|
|
Advertisement
|
|
| |
| Blog |
|
|
| Castle Party |
|
Today's issue is very short. The conference ends at around noon, and we want to be sure we can still use the Internet connection here to get the issue to you. The Castle Party was OK. Very nice, very old large room, good cuisine and nice company. The ISC2004 Awards were presented and there was music by a Heidelberg group, with non-typical Heidelberg instruments, like an electric piano, an accordeon, and a 3 metre long waldhorn. This morning I tried to visit the remaining booths at the exhibition. This included Quadrics, displaying their QSnetII, and several integrators from Germany, CADAC and CPI. Clearspeed had two 0.25 Tflop/s machines (32-bit, peak) on display using their attached processor. Fujitsu Siemens showed the newest version of the hpcLine. (AE)
Read further... |
|
|
|
| Be honest |
|
About ten years ago, David Bailey wrote an essay about "twelve ways to fool the masses", warning for misunderstandings about parallel computing. In Heidelberg he revisited that essay with an add-on to the title "back to the future". Collective misunderstandings are common in science. Some are even firmly based on experiments; be it sloppy, and biased experiments. In the hype around parallel computing in the early nineties, successes were exaggerated. An example was that 8,192 CPU's results were scaled to 65.536 CPU's and the latter were presented as measured results. Bailey has about twenty of these type of papers at home. But no reason to try to break into his home. They are safely locked away, with copies on secure other places. New methods of fooling the masses: run codes dozens of times, only use the best results. Or : "scalability" is defined as a succesful execution on a large number of CPU's, regardles of performance. Today, the Grid shows similar "methods": all supercomputer computation will be done on Grids, is one that is heard a lot, another: a computational grid has a capacity greater than the constituent systems", which is not quite true either. What should we do? Try to be honest, it is easy to fool yourself too. According to Bailey the only way is to use well-designed, community designed, rigourous benchmarks. (AE)
|
|
|
|
Advertisement
|
|
Advertisement
|
|
| |
| Applications |
|
|
| TENT and DataFinder systems allow complex numerical simulations at the German Aerospace Centre |
|
The German Aerospace Centre does research in astronautics, aeronautics, energy and transportation. There are 5000 employees at eight major sites and about thirty institutes. The central DLR facility provides software engineering at SISTEC, the centre for simulation and software technology. Rolf Hempel described the TENT and DataFinder systems which allow researchers at the centre to run complex simulations.
Read further... |
|
|
|
| ISC2004 AMD Award honours TeraGyroid experiment |
|
Stephen Pickles, one of the winners of the ISC2004 AMD Award, presented the TeraGyroid experiment. The project is funded by EPSRC in the UK and NSF in the USA to join the UK e-Science Grid and the U.S. TeraGrid. It constitutes an application from RealityGrid, a three-and-a-half year UK e-Science project including work exhibited at SC 2003 and SC Global in November 2003. It has received thumbs up from TeraGrid in mid-September of last year and funding from EPSRC approved later. The main objective was to deliver high impact science which it would not be possible to perform without the combined resources of the U.S. and UK Grids.
Read further... |
|
|
|
|
|
|
|
|
EnterTheGrid - Primeur
James Stewartstraat 248
1325 JN Almere
The Netherlands
http://EnterTheGrid.com
mailto:primeur@hoise.com
|
|
|
|
|
|
