NIC Symposium 2014 - Proceedings

As part of a well-established and very useful tradition, the John von Neumann Institute for Computing (NIC) has been holding NIC Symposia biennially since 2001 in order to present highlights of the research that is only possible by using extensively the resources provided by the NIC and the Jülich Supercomputing Centre (JSC). In June 2013, the massively parallel machine JUQUEEN of the JSC, which contains almost half a million processors, reached the status of "No. 1 in Europe" on the "TOP 500" list of supercomputers. However, the main achievement of the JSC is not just an excellent score in this type of ranking, but the fact that the excellent hardware performance actually does enable really outstanding research of the highest quality in research fields encompassing astrophysics, computational biology and biophysics, chemistry, elementary particle physics, materials science, condensed matter, computational soft matter science, earth and environment, computer science and numerical mathematics, fluid mechanics and plasma physics. There are several ingredients which all play a pivotal role to continuously maintain the top standard of supercomputer based research at the NIC/JSC. In particular, the user-friendly and very helpful staff of the JSC provides fast and easy access to the machines and creates ample opportunities for well-targeted training. One must keep in mind that many of the computational codes (which often are based on code developments extending over decades and hence sometimes are very complex and sophisticated) are not immediately suited for a massively parallel computer architecture. In some cases the training and advice provided by the JSC staff allows an adaptation of these codes. In many cases the scientific problem is intrinsically in conflict with a massively parallel execution of the code, however: for these problems it is of crucial importance that the JSC has a dual hardware strategy, providing with the system JUROPA (which has a cluster architecture with faster, but less energy efficient, processors and only moderate parallelism) a "general purpose supercomputer". This indeed has served the needs of about 90% of all the (several hundred!) research projects, that were carried out at the JSC in the two years since the last NIC symposium