JARA-HPC: Computer Simulation Using High-Performance ComputersPeter Winandy
Computer simulations have become the third pillar of research next to theory and experiment. They open up new scientific perspectives that for physical-technical or financial reasons have so far been denied. Researchers of nearly all scientific fields can benefit from simulations: from medical sciences to engineering, information technology and material sciences up to environmental and energy research. Many simulations, however, require the enormous processing power of supercomputers.
High-Performance Computing in the Service of Science and Research
Scientists of the section JARA-HPC– short of High-Performance Computing– unite the specialist know-how of highly parallel computing on supercomputers with the respective special knowledge of physicians, engineers, and other scientific researchers.
In doing so they considerably contribute to making full use of the opportunities computer simulations offer to address current scientific issues. JARA-HPC scientists are part of a unique organizational structure comprising socalled Simulation Laboratories – SimLabs, Cross-Sectional Groups – CSGs, and JADE.
Research into Application
Among other achievements, computer simulations helped researchers of JARA-HPC to improve blood pumps that support the heart of seriously ill patients, and to create a "virtual brain" with which brain scientists can now better investigate and understand interconnections of the real brain.
They have calculated unusual magnetic and electronic properties of materials, leading to the manufacture of novel electronic components or data storage systems on the basis of the knowledge acquired through simulations.
Furthermore, JARA-HPC scientists have developed computer simulations with which the effects of human activities on the biodiversity of meadows and grasslands in the Nationalpark Eifel can be predicted.
Researchers of the section HPC have access to supercomputers that internationally excel in performance, in particular via Jülich Supercomputing Centre (JSC). They are developing application software keeping pace with the constant performance increase of high-performance computers.
Scientists of the section JARA-HPC combine the necessary modeling approaches and numerical methods for massively parallel computing on supercomputers in the SimLabs.
There are currently three SimLabs in different scientific disciplines in which users are provided targeted support in the development of new simulation applications and the optimization of previously existing simulation applications:
- SimLab Highly Scalable Fluids & Solids Engineering: Research and support for users in structural and fluid mechanics for the use of massively parallel systems
- SimLab ab initio Methods in Chemistry and Physics: Research and support for users for ab-initio simulations in chemistry, physics, and the natural sciences
- SimLab Neuroscience: Development of models and databases for large network simulations of the brain
The work in the SimLabs is supported by cross-sectional groups. These teams include mathematicians and computer scientists, whose methodological know-how benefits all users of high-performance computers.
- CSG Immersive Visualization: Development of algorithms for immersive, interactive, and spatially distributed and cooperative visualization
- CSG Parallel Efficiency: Development of algorithms to analyze and increase the performance of parallel programs
The JARA HPC Partition is the merger of shares of HPC systems in Jülich – JUQUEEN and JURECA – and Aachen – RWTH Compute Cluster. Main reasons to set up this Partition in 2012 were the guaranteed amount of computing time for projects from the two regional partners on latest HPC architectures, the reduced but highly efficient assignment procedure that fosters scientific excellence, and the requirements of the funding bodies BMBF and MIWF to arrange for larger and more efficiently used HPC infrastructures.
The computing capacity of the JARA HPC Partition is continuously being expanded in order to also offer excellent conditions for work and research with and in the simulation sciences in the future.
The JADE – Jülich Aachen Data Exchange tries to establish a flexible and scalable data management tool that addresses the specific needs and requirements of domain scientists to exchange data between Forschungszentrum Jülich and RWTH Aachen in a convenient and efficient way.
The seed funds for funding promising project not yet capable of receiving external funding proved their importance and value in Institutional Stratey I.
For this reason JARA provided another 750,000 euros for seed fund projects in Institutional Strategy II. These seed funds could be jointly acquired by scientists at RWTH Aachen and Forschungszentrum Jülich in the spring of 2014.
A total of 23 project proposals with a requested funding volume of 5.1 million euros were evaluated by external reviewers. Five of the best proposals received funding volume ranging from 125,000 to 250,000 euros for a duration of 18 to 24 months.
|The following Seed Fund projects were funded in JARA-HPC in the ZUK II :|
|NESTML - A modelling language for spiking neuron and synapse models for NEST|
|Hybrid parallelism for the Jülich DFT code FLEUR|
|Interactive Volume Rendering of Massive Data on the Blue Gene Active Storage Architecture|
|Development of a DEM process model for HPC of vibratory finishing|