"Given the complexity of three-dimensional flow phenomena, it is not always easy to predict how our projects are going to evolve", explained Thomas Damrau, IT Manager at the BAW Institue. "Our employees regularly need to make significant modifications to models and the various parameters, and even may have to change the source code. That involves testing and validating the computing code that has been changed, so they need direct, rapid access to the Data Center. This means that the BAW Institute's supercomputer has to constantly adapt to evolving needs."
The BAW Institute's supercomputer consists of a cluster of seven bullx blade chassis with 126 processing nodes, representing a total of 2,016 cores. The processing nodes are connected via a high-speed InfiniBand QDR interconnect network, with an Ethernet network serving users and the network administrator. The blades were specifically designed by Bull intensive computing applications and are based on the latest generation of Intel Xeon processors (Sandy Bridge-EP), Quick Path Interconnect, DDR3 memory and QDR Infiniband technologies. The Panasas PAS12 storage systems connected via InfiniBand deliver total capacity of 100 TB. Thanks to an innovative water-based cooling system, the supercomputer consumes very little energy. In addition, the heat it produces is collected in the system so it can easily be reused by the BAW Institute, for example for space heating.
A key feature of the configuration is the potential it offers to managers at the BAW Institute to monitor the system using the scVENUS management tools from science + computing AG, a Bull subsidiary. Another advantage is that the existing servers at the BAW Institute can also be administered using the same tool.
"scVENUS - which is automatically installed by the operating system - allows to be configured the cluster environment, the necessary applications to be installed and optimizes task planning", explained Thomas Weselowski, Director of Extreme Computing, Bull Germany. scVENUS offers a wide choice of standard, proven HPC software, and are well suited to requirements, library administration tools and compilers. "This means we can provide the optimum response to users' demands, while at the same time improving the price/performance ratio by 10-15%", added Thomas Weselowski.