Max Planck Institute Halle calls on MEGWARE to expand its cluster system
The Max Planck Institute (MPI) Halle has expanded its HPC system ‘Cruncher.’ The new cluster segment was successfully put into operation using the MEGWARE compute platform SlideSX® and Intel Xeon ‘Haswell’ processors. The project – from the planning phase and production of the cluster through the installation phase and handover of the entire system on site – was managed and completed by HPC specialists MEGWARE.
MPI Halle’s need to expand its HPC cluster arose from the CPU-intensive tests that it needs to perform in its data center to solve nanostructure physics problems. The cluster was expanded to include 148 MEGWARE MiriQuid® computing nodes, an MDS server (both dual socket systems), and several storage servers based on Intel Haswell CPUs. In addition to SlideSX®, two further MEGWARE products designed specifically for the HPC market were used. ClustSafe®, for the electrical connection of the computing nodes and storage servers, and ClustWare®, for facilitating the monitoring and control of the HPC cluster, round off the perfectly integrated HPC solution.
The new partition therefore consists of almost 3,600 computational modules, with a theoretical peak performance of 144 teraFLOPS. Cruncher has been enhanced to approx. 12,000 cores. The size of the central storage system, based on the parallel cluster file system BeeGFS, is now around 500 TB.