The German federal scientific computing network –
D-Grid – now features a Bull NovaScale® supercomputer
Hosted by the Leibniz Universität Hannover, it provides German higher education and research centers with new levels of computing power
The Regional Computer Center for Lower Saxony (RRZN - Regionales Rechenzentrum für Niedersachsen) has invested in a Bull NovaScale supercomputer, putting new levels of computing power at the disposal of the D-Grid community at federal level.
RRZN is one component in the intra- and inter-regional networks that have been created to supply German higher education establishments and research centers with state-of-the-art high performance computing resources. In order to achieve this, the Leibniz Universität Hannover is now hosting the NovaScale supercomputer.
“We are very pleased that RRZN now ranks among our major high-performance computing (HPC) customers. This new achievement confirms the progress our computing solutions are making in European and worldwide research. The choice made by RRZN, following is the footsteps of those made by other major research centers in France, Germany, Ireland, Spain,and the United Kingdom, supports our strategy to provide a highly innovative platform for HPC which has the capacity to combine the increased levels of power with an extremely compact, low-consumption and easily administered solution,” declared Benoît Hallez, Director of Bull’s HPC business.
63 servers and 2.5 Teraflops in just two racks
To meet the requirements of the European tender set out by RRZN, Bull proposed a NovaScale cluster comprising 60 1U computing nodes equipped with Intel® Xeon® dual-core processors, and two 2U system administration nodes (NovaScale R460), also equipped with dual-core processors. This solution is completed by an image server (NovaScale R440), giving a total of 63 servers that are integrated into just two 40U racks.
The NovaScale supercomputer has achieved the best benchmark test results, and offered a very competitive price/performance ratio. It delivers a peak performance of 2.5 Teraflops, and fulfills a particularly demanding set of specification requirements, notably:
• The need to fit in a very limited floor space
• Moderate thermal dissipation, requiring air cooling only
• A cluster architecture of SMP servers running under Linux®
• Extended set of remote administration functionalities.
Available to the whole D-Grid community, the system has been in production since January 2007, having passed the benchmark compliance tests with flying colors and successfully undergone 14 days of intensive stress tests.
About the D-Grid community
The D-Grid initiative sets out to deploy integrated networks for sharing Grid resources enabling the computing of huge volumes of data, thanks to a high level of virtualization when it comes to managing resources and data.
D-Grid was first put forward in 2003 on the initiative of a number of German researchers and research institutes, who jointly published a common strategic declaration calling for, in particular, the development of grid network technology to make intensive computing available to the German scientific community. In 2004, the initiative received the support of the German Ministry for Research and Education (the BMBF, or Bundes Ministerium für Bildung und Forschung).
Work to deploy the new infrastructure began in 2005, with the creation of the D-Grid integration project and start-up of six community grids designed especially to cater for high-energy physics, astrophysics, medicine and life sciences, climatology and engineering.