HPC Center

Significant Infrastructure / Major Items of Technical Equipment

Dragon`s-Rock Cluster
© Fraunhofer SCAI
Dragon`s-Rock Cluster

SCAI operates a modern, powerful and diverse IT infrastructure including several High Performance Computing (HPC) and Big Data clusters. This includes:

  • an HPC-HPDA cluster with 128 nodes (4096 Skylake EP CPU cores, Omnipath fabric with fully bisectional 100 Gbit bandwidth in fat-tree topology) plus 16 nodes with additional local PCIe NVMe for data analytics,
  • an HPC cluster with ~300 nodes (4096 Ivy Bridge CPU cores, 640 AMD cores on big memory machine, FDR Infiniband 56GBit interconnect),
  • a shared parallel distributed file system of ~480 TB capacity distributed over 13 storage and metadata nodes connected with Infiniband and Omnipath,
  • dedicated nodes with NVIDIA Tesla GPUs (including 12x Tesla A100 & 8x Tesla V100) and Intel KNL, FPGAs,
  • a data-analytics cluster based on distributed Ceph object storage with ~400TB capacity and various virtualization hosts for running Hadoop, Accumulo, Cassandra,
  • a Kubernetes Cluster for production hosting of microservices including a Ceph-based storage system with more than 1PB of capacity.
  • multiple in-house private Cloud infrastructures based on OpenStack (separated for production services and research projects)

As a member of the Fraunhofer Competence Network Quantum Computing, SCAI has access to a quantum computer operated by IBM at its location in Ehningen, Baden-Württemberg.

Service Portfolio

© Fraunhofer SCAI

SCAI IT-S offers support and services for external customers and project partners

Our competences:

  • Performance benchmarking and optimization
  • Automation of application workflows, distribution of work and data in parallel and/or distributed environments such as HPC clusters and Clouds
  • Consulting and custom development of Middleware for service integration, REST services and APIs, distributed Authentication and Authorization
  • Support for using HPC and Cloud infrastructures efficiently for arbitrary computations such as simulations or data analysis
  • HPC and Cloud related trainings (e.g. Slurm, MPI, CUDA, etc.)
  • HPC support for service providers
  • Deployment and Operations consulting for your in-house HPC or Cloud systems

Cloud Computing

Over the last years, Cloud Computing has evolved to be the by-far leading approach to use, deploy, and operate infrastructure and software services. The benefits that most customers appreciate are the greater flexibility, lower up-front investment costs and lower operational expenditure in the medium term though resource usage that is adapted to the respective demand. These benefits apply both to providers and users of software and services. When talking about Clouds, we usually differentiate between the different levels of service provision: Infrastructure as a Service (IaaS), Platform as a Service (PaaS) und Software as a Service (SaaS), with a decreasing  dependency of the user (or customer) regarding the hardware actually used by the Cloud service provider. Furthermore, Clouds are distinguished by their deployment model, for example in so-called public Clouds (commercial offerings), in-house or private Clouds (which may be operated by the local IT department or a commercial provider), hybrid Clouds that combine both types, and community Clouds that are operated by and for a dedicated community. Especially due to the strong growth and wide adoption of Internet of Things (IoT), the Edge Cloud or more generally Edge and Fog computing have gained significant importance for businesses and research.

All of these approaches or realisations come with their own share of challenges for adoption, technical realisation, secure and efficient operations, etc. both for providers and users. Amongst others, these might concern Scalability and Performance, Data Management, automated distribution of work, Interoperability with and Integration with local legacy infrastructure and services, Security and Privacy, Data Protection, Service Level Agreements, Software Licence Management. In several national and European funded projects Fraunhofer SCAI has developed approaches addressing these matters and has acquired expertise, which we offer to our customers and partners.

SCAI is a member of and Cloud Computing provider in European Open Science Cloud through the participation of the German e-Infrastructure initiative NGI-DE in EGI-ACE, providing a Cloud Infrastructure as part of the EGI Federated Cloud (FedCloud).