Report on the SC2000 ConferenceUtrecht 30 November 2000 In this report we give an impression of the SC2000 Conference held in Dallas, Texas, USA, on 4-10 November. We discuss some trends and new developments as noted at the conference and the associated exhibition.
Dallas, Texas, was hosting SC2000 IEEE/ACM conference on Supercomputing and Communication. This biggest yearly event in the HPCN area draws about 5000 people in all since many years because there is not only an extensive technical program but also a 3-day exhibition where all vendors, large and small, can show off their latest products and services. In addition, there is the research exhibition in which universities and research labs show their latest achievements. The SCwxyz series of conferences always has a large variety of activities apart from the regular refereed papers. We give a few statistics: 24 Tutorials (on 2 days prior to the conference proper), 7 panels, 29 poster presentations, 7 panel sessions, 20 Birds-of-a-Feather meetings, 4 State-of-the Field talks, 12 ``Masterworks'' invited talks, 95 commercial and 60 research exhibits, and, new this year, the HPC games. In this contest the players were requested to build a cluster for under US$ 10.000 and to perform a predefined benchmark exercising CPU, network, and I/O of the cluster. A new feature of the plenary sessions (the Keynote address and the State-of-the Field talks) was that were webcasted to enable a much larger audience to follow them.
Each of the SC conference has over the years had its own dynamics and
atmosphere depending on the main focal points for research and what was
emerging in the market. In this respect SC2000 has been one of the most
interesting conferences in the last few years. Not only it is evident
that cluster have made themselves a definitive place on the HPC stage,
also two new topics begin evolving with a quite large momentum: grid
computing and cycle harvesting (``Megacomputing''). It is too early to
predict what the impact of these two new fields will be but they will
certainly alter the HPC scene. These developments are enhanced by the
fast growth in bandwidth and storage density that presently grow with a
speed that is comparable or in some areas even larger than that of the
processor speed.
Furthermore, genomics has become a prime target for HPC which requires
new techniques and perhaps even new machine configurations to accommodate
these requirements. Genomics may be a factor in focussing grid-oriented
computing as the necessary functionalities in this field can differ so
much that employing different systems for different tasks comes as a
natural thought.
All these new developments gave the conference an air of expectancy that
extremely interesting work in many areas can be done in the very near
future together with a feeling that the HPC field is in a state of
change of which the implications can not yet be fully understood. It
makes one very curious about what will be presented at SC2001 that will
be held 10-16 November in Denver, Colorado, next year.
Aad van der Steen |
 |