High Performance Computing

The group administers different Linux clusters mainly based on Blade technology. We maintain 27 Bladecenters with 5220 CPU cores for the theory groups. To the users, they all look alike as they use the same set of compilers, libraries and batch systems. On all nodes Linux is used as Operating System. 40 kW of energy savings in recent years were needed to host the compute cluster of the newly established theory group of Ali Alavi locally. This machine consists of 144 computing nodes with 20 CPU cores and 128GB main memory each, thus in total 2880 cores and 18 TB main memory. Under full load the Alavi Cluster consumes 45 kW. Part of the cluster is a IBM GSS storage appliance hosting a 450 TB GPFS parallel filesystem. This filesystem is exported over a dedicated 10GbE line to the Max Planck Computing center Garching (RZG) where the group uses the Hydra supercomputer for production runs.