Data Science & Computing Lab

Founded in Sept. 2020, the Data Science & Computing Lab (DSC Lab) strives for scientific discovery and innovation through data-based and computing-intensive methodologies. The Lab's research addresses scientific questions, where

• additional insight and understanding of complex phenomena can be gained by modelling and learning from data 
• state-of-the-art parallel algorithms and high-performance computing (HPC) facilitate the efficient simulation of very detailed and fine-grained numerical models to advance scientific discovery in areas, that are hard or even impossible to be discovered without computing techniques.

With its expertise in data science and HPC as well as its hardware resources, the DSC Lab supports all computing-related research activities at the Institute for Mathematics and Computer-Based Simulation as well as aids the transfer of research findings into application codes of our academic and industrial partners.

HPC Hardware

For small- to mid-scale parallel computations, we operate a high-performance computer (Linux cluster) for parallel computations. The key technical specifications of our cluster are:

• Linux cluster of compute nodes with multi-core CPUs
• 480 cores in 20 Intel Xeon Skylake-SP nodes (2x12 core, AVX512), 16x 196 GB DDR4 RAM per node & 4x 768 GB DDR4 RAM per node
• 832 cores in 16 Intel Xeon Cascadelake nodes (2x26 core, AVX512), 4x 384 GB DDR4 Ram per node
• Melllanox Infiniband Network
• rack cabinet with integrated water cooling

We also operate a network of multi-core testing machines to support CI/CD software development at IMCS.

HPC Software

The DSC Lab actively participates in the development of several parallel research codes, open-source math libraries, and software environments:

• Baci: Baci is an interdisciplinary software project for massively parallel simulations of coupled multi-physics phenomena. It allows for the analysis of solid dynamics, incompressible fluid flow, convection-diffusion-reaction, and photoacoustics problems and their interactions. Specialized reduced order models for cardiovascular applications as well as capabilities for inverse analysis, optimization, and uncertainty quantification complement its capabilities (in collaboration with LnM@TUM, IMCS@UniBwM, and HZG).
• Trilinos: The Trilinos Project is an open-source effort to develop algorithms and enabling technologies within an object-oriented software framework for the solution of large-scale, complex multi-physics engineering and scientific problems.
• MueLu - the next-generation multigrid framework within Trilinos: MueLu is a flexible and scalable high-performance multigrid solver library. It provides a variety of multigrid algorithms for problems ranging from Poisson-like operators over elasticity, convection-diffusion, and Navier–Stokes to Maxwell’s equations.
• Spack: the flexible package manager for HPC, Linux, and macOS

Teaching

The DSC Lab is involved in teaching at the B.Sc. and M.Sc. level in study programs such as "Civil Engineering and Environmental Sciences" or "Mathematical Engineering". The DSC Lab contributes to these courses:

• Introduction to Programming
• Advanced Topics in Numerical Methods
• Introduction to the Finite Element Method
• Non-linear Finite Elements
• Modelling of Uncertainties and Data

Moreover, the DSC Lab facilitates student theses through academic guidance as well as access to computing software and hardware.

Industrial Partners

The DSC Lab cooperates with selected industrial partners in a number of research projects and efforts towards technology transfer ("from research into application"). Within the scope of these projects, our industrial partners benefit not only from the Lab's bundled expertise in software development, data science and scientific computing, but also from its excellent technical equipment. Have a look at our list of current and previous industrial partners as well as opportunities for collaboration!