Workshop

Amsterdam, The Netherlands

http://www.wins.uva.nl/events/HPCN2000/

Dr. Patrick Aerts is the director of the Dutch National Computing Facilities Foundation.

Dr. Mark Portney received a BS in Math and Physics at UCLA in 1976, MS in Physics (theoretical solid state) at Arizona State University (ASU) in 1978, and PhD (theoretical particle physics) at ASU in 1983. After postdoc work in Geophysics at ASU, he worked at Texaco in Houston from 1984 to 1993. He was involved with all aspects of seismic data processing, and worked on a variety of computers. He moved to SGI's Technology Center in 1993 in order to work on SGI multiprocessing computer systems. His responsibilities since then have been benchmarking and performance engineering for technical computing. He is especially interested in high-performance I/O.

Dr. Aad van der Steen studied mathematics at the Delft University of Technology while providing computational support at the Interuniversity Reactor Institute in Delft. After a short stay as a mathematical modeller at the Environmental Ministry he was employed at the Computing Centre of the Utrecht University and as an advisor to NCF for High Performance Computing. He obtained his PhD. from the Computational Physics Department of the Utrecht University on the subject of benchmarking High Performance computers. He currently works for this department and for the NCF.

Dr. Harrie Weinans studied mechanical engineering at University of Twente (degree 1986). PhD in medical science from University of Nijmegen (degree 1991).From 1994-1996 employed at Rush Medical Center, Chicago. From 1997 until now employed at Erasmus University Rotterdam. Research interest: the interaction between the biological and mechanical processes in bone tissue in relation to the fixation of orthopaedic implants and bone fracture risk.

Arthur Veldman obtained a Ph.D. in Applied Mathematics from the University of Groningen. In 1977 he joined the National Aerospace laboratory NLR in Amsterdam, where he was involved in various projects in the area of computational aero- and hydrodynamics. In addition, between 1984 and 1990 he was part-time professor of CFD at Delft University of Technology. In 1990 he returned to Groningen, where he now occupies the chair in Computational Mechanics.

To be announced.

The Dutch National Computer Facilities Foundation (NCF) commissioned a new computer system to replace a Cray C90 that is used nationally by many universities and research institutions. Part of this process was an extensive benchmark to evaluate the performance of the candidate systems.

We describe the structure and methodology of this benchmark and present some of the results that may be disclosed. The benchmark results have, at least for a major part, lead to the procurement of the 1 Tflop/s SGI SN-1 system of which the installation will be completed in November 2000.

Short overview of numerical simulations of various aspects of bone tissue. The internal morphology of bone consists of an irregular spongy (foam) like structure. This has important consequences for the load bearing function of bone. Numerical simulations of the load distribution through the architecture were performed. This elucidates how effective the structure can withstand loading. In addition it can be simulated how bone is continuously being resorbed and replaced by cells. The interaction between the cellular processes and the load distribution will be discussed. This research can be used to understand the mechanics of bone which has important implications in diseases such as osteoporosis and arthroses.

Turbulent flows possess many, seemingly chaotic, details with small length- and time scales. Much research is going on into the character of turbulence: the main question is when and how small-scale turbulence features influence large-scale flow behaviour.

Computer simulations, better than experiment, are able to provide detailed insight into turbulent phenomena, at least in principle … However, the required computer power is large, exhausting every supercomputer. Fortunately the latter become more and more powerful, and a comparable progress in algorithmic efficiency can be mentioned: together a factor of a thousand per decade is gained.

The "amount" of turbulence in a flow is mainly determined by the Reynolds number, which is a measure for the ratio between convective inertial forces and diffusive viscous forces. Increase of the Reynolds number corresponds with an increase of the ratio between the large an the small scales in the flow, and hence an increase in the computational complexity of the simulation: the latter scales with approximately the third power of the Reynolds number.

For various industrial flows, with Reynolds numbers around tenthousand, detailed turbulence simulations are currently possible. But the flow around a car, for instance, has a Reynolds number of one million, and hence is a million times more expensive. A great computational challenge for the coming decade!

SARA

NCF/NWO

Last updated on April 27th, 2000

Created by:

E-mail: annef@wins.uva.nl