SiVeGCS

Sicherstellung der weiteren Verfügbarkeit der Supercomputing-Ressourcen des GCS
Keyvisual image main

SiVeGCS coordinates and ensures the availability of HPC resources of the Gauss Centre for Supercomputing, addressing issues related to funding, operation, training, and user support across Germany's national HPC infrastructure.

SiVeGCS provides critical funding for Germany's three national supercomputing centers within the Gauss Centre for Supercomputing (the High-Performance Computing Center Stuttgart, the Jülich Supercomputing Centre, and the Leibniz Supercomputing Centre). In addition to supporting the development and operation of GCS's supercomputing systems, the project enhances the centers' ability to provide users with high-level support services, in particular for enhancing the efficiency of their applications. The following are some key components.

User support

Today's supercomputers are not just expensive but also consume huge amounts of energy. For example, operation of HLRS’s current flagship system, Hawk, requires approximately the same amount of energy as a city of 24,000 residents! This means it is critical that resources be used as efficiently as possible; i.e., achieving as much scientific progress as possible with respect to runtime and energy consumed.

Typically, one effective way to address this challenge is to optimize the performance of applications that run on HPC systems. Under the auspices of the SiVeGCS-MWK project, BMBF and the State of Baden-Württemberg Ministry of Science, Research and Art (MWK) therefore provide HLRS with funding not only to purchase new hardware but also to hire user support staff at the center. At HLRS, scientific staff works closely with our system users to assist them in optimizing their applications. Similar user support capabilities also exist at other centers within the Gauss Centre for Supercomputing, with support from the corresponding state government.

Runtime

01. January 2017 -
31. December 2025

Website
Funding

BMBF & MWK Baden-Württemberg

To optimize applications, we start with a detailed analysis of representative application runs. Based on this analysis, the HLRS user support staff develops ideas for optimization. In some cases we implement them in close collaboration with the application user. In others we provide recommendations to system users that they can implement themselves. The scope of such optimizations is not limited to increasing scalability but also includes improving node level performance, selecting algorithms that are well suited for the respective architecture, and improving I/O.

Although the performance of some applications can be improved by applying known optimization methods, HLRS user support staff members also develop new optimization methods when no suitable method is yet available. Once developed, such new methods can also be applied to other codes that might in the future benefit from similar optimization strategies. In some cases, these developments can also be published in scientific papers written by HLRS staff scientists and the system users involved.

Further tasks of the work package are to prepare applications for future systems, to develop and implement new modes of operation, and to support academic communities and industry in the use of supercomputer resources.

Training

HLRS provides and continues to develop HPC training activities in ways that are consistent across the three centers comprising the Gauss Centre for Supercomputing. This includes, but is not limited to, updating content continually to address new technological developments, customers’ needs, and the scientific focus of each center. In addition to new course content, SiVeGCS also enables the development of new course formats, including online courses, hackathons, or the use of Jupyter Notebooks and compute clouds.

Industry collaboration

Making the technologies and know-how gathered at GCS available to industry — especially to small and medium-sized enterprises (SMEs) — can support the competiveness of the local economy on the international stage, including indirectly helping to secure jobs. SiVeGCS coordinates the implementation of this basic idea at the three GCS centers and organizes informational events, multiplying the use of future-oriented HPC technologies across German industry.

Project partners

  • High-Performance Computing Center Stuttgart
  • Jülich Supercomputing Centre
  • Leibniz Supercomputing Centre