Bernd Wiebelt (Computing Center, University of Freiburg) | High Performance Neuro Computing

Abstract

In the past decade, High Performance Computing (HPC) has developed into a key technology for brain research. In particular, the rapidly growing field of Computational Neuroscience has seen exciting new developments, both in multi-variate data analysis and in large-scale model network simulations. Available software packages scale in performance from the most powerful supercomputers to small compute clusters for local work groups, the latter often being the point of entry for aspiring young scientists. However, these mini-clusters come with strings attached: Find the money, acquire the hardware, provide software and human resources to keep everything afloat. And after 3-5 years, start from scratch because everything is outdated. An alternative approach has been developed by the state of Baden-Württemberg by way of its bwHPC initiative. The bwUniCluster in Karlsruhe and the upcoming bwForCluster ENM in Freiburg are Supercomputer class HPC clusters shared by scientists from the whole state. The bwForCluster ENM is especially targeted towards the Computational Neuroscience community. The idea is to preserve the benefits of mini-clusters, such as straightforward access, gradual learning curve and benevolent error tolerance. At the same time, by joining forces, we were able to acquire far more resources and make much better use of them.