New technologies and trends for future HPC systems are monitored, evaluated, and integrated into the network’s activities, particularly in the context of AI. The goal is to consolidate the centers’ existing activities, conduct complex evaluations collaboratively, and document the results in a way that benefits all locations and, where possible, leads to publication.
Key areas of focus include:
- Hardware architectures for new use cases in HPC, e.g., AI accelerators
- Unified memory architectures for CPU/GPU to simplify the use, programming, and porting of applications and workloads
- Energy-efficient computing with flexible accelerator architectures
- Alternatives to x86-based systems for energy-efficient HPC
In addition, thematic specializations are being established within the network to pool knowledge and offer cross-regional consulting and training. The focus on specific areas and specialization follows these criteria:
- Discipline-oriented (e.g., engineering, computer science, physics, chemistry, life sciences, digital humanities),
- Method-oriented (e.g., parallelization methods and paradigms, debugging, performance analysis, performance tuning, machine learning, big data analytics, and high-performance visualization),
- Operations-oriented (e.g., batch and scheduling systems, system monitoring and planning, performance and utilization monitoring).
Activities are evaluated through documentation, user surveys, and the tracking of supported research projects and resulting publications.