Fraunhofer SCAI

The Fraunhofer Institute for Algorithms and Scientific Computing (SCAI) conducts research and offers services in the research areas of numerical software, simulation engineering, optimization, distributed computing, bioinformatics and virtual material design. Fraunhofer SCAI has relevant experience that ranges from research on parallel numerical algorithms over software porting and service integration, such as AA and accounting integration, to professional benchmarking for customers from industry and SMEs. Commercial software products of the institute are running on a variety of systems, especially HPC clusters. The institute was one of the first CUDA research centres in Germany, developing and optimizing CUDA based solutions for industry and SME partners. Through its unique position between research and industry, SCAI is able to provide its SME contacts to the activities of the project. SCAI will offer its broad experience in HPC related research and product implementation activities through consultancy and development of best practices.

• Code optimizing & Software development
• Performance benchmarking
• HPC support for service provider
• Assistance using simulation codes
• Training (e.g. CUDA, on request FPGA related trainings)
• Operation and installation consulting
• Additional: Consulting for data analytics applications on HPC systems

No. of systems:

2 (Drachenfels and Loewenburg)

Architecture(s):

Drachenfels: Mixed HPC cluster with GPU and Intel PHI with 300 Nodes, FDR Infiniband

Loewenburg: HPC-Cluster with 144 Nodes, OMNI-Path-Interconnect

Storage:

• Parallel File System – 200 TB
• Long Term Storage – 300 TB

Testsystems for Accelerators:

TESLA, Xeon Phi

HPC applications and software available:

Development Tools: different compiler and performance tools (not limited to academic), Microsoft HPC

ISV-Codes: e.g. Abaqus, LS-DYNA, Star-CCM+, Fluent ( licensees need to be negotiated)

Opensource: e.g. OpenFOAM, OpenFOAM with commercial AMG solver, GAMESS, Chromacs, CP2K, AMBER (partly only academic)

AOB: solutions for High Throughput Computing (HTC) e.g Moab Nitro

Comment: customer dependent software will be installed by request

Access

portal or interactive

Cloud Resources

• 16 Node Openstack federated cloud resources (see here)
• 8 Node Opennebula secure cloud resources (secure infrastructure for health Care analytics see here)

• Multiphysics with legal ISV code (MPCCI, commercial)
• Parallel algebraic Multigrid solver (SAMG, commercial)
• Uncertainty in engineering design Desparo, commercial)
• MD for material design (Tremolo-X, Commercial)
• Library for Accelerated Mathematical Applications (LAMA, Open Source)