Center for Computational Sciences and Simulation (CCSS)
Research
High performance computing solves problems that are too complex for classical methods. This is made possible by initially modifying (discretizing) the relevant equations so that a computer algorithm can find a solution iteratively. In this process, the members of the CCSS are responsible for development of the algorithms, implementation in software, and use of the programs.
The phenomena these programs are used to explore often take place on different scales (e. g. from a drop of fuel to a gas turbine), and the influence of the small scales on the simulation as a whole must then be modelled using “closure models” or “scale transition conditions”. These modelling theories are similarly developed by members of the CCSS and tested in complex simulations.
Members of the CCSS from biology, chemistry, computer science, engineering sciences, mathematics and physics often work with problems that require tremendous computing capacity to solve – executing the programs sequentially would often take decades. It is therefore essential that different computations are performed simultaneously – a procedure referred to as parallel computing. This is only possible with efficient, often complex parallelization strategies as well as suitable parallel computers. The UDE’s high performance computer, a Cray XT6m, has 4128 cores and will be joined in 2016 by a NEC system with 13,248 cores (see below).
This high performance computer is a key instrument in gaining external funding, attracting outstanding employees and conducting research projects with an international profile. At the same time, the high performance computer enables the researchers at the UDE to develop their algorithms for supercomputers one to two orders of magnitude more powerful (e.g. Jülich, Munich, Stuttgart). This has meant that various research groups now have the capability to undertake top-level research that requires many times greater computing capacity than at the UDE. It is precisely in helping the UDE researchers to this capability that the CCSS and ZIM support team plays a decisive role (https://www.uni-due.de/ccss/sc_support.php).