ChEESE-2P

ChEESE, the Centre of Excellence for Exascale in Solid Earth, aims to become a hub for HPC software within the solid earth community. ChEESE is addressing multiple challenges posed by geohazards by developing improved methods for their accurate prediction using exascale computers.

The scientific ambition of ChEESE-2P is to prepare eleven community flagship codes to address 12 domain-specific exascale computational challenges. The project expands upon areas covered during the first implementation phase (computational seismology, magnetohydrodynamics, physical volcanology, and tsunamis) to include two additional disciplines (geodynamics and modeling of glacier hazards). Using a co-design process involving mini-apps, codes will be optimized in terms of their performance on different types of accelerators, scalability, deployment, containerization, and portability. ChEESE-2P will investigate these issues both for current pre-exascale systems and for future hardware architectures emerging from the EuroHPC Pilots. Optimization will also consider heterogeneous single-node and multi-node performance, and will include continuous efficiency monitoring using the Performance Optimization and Productivity (POP) metrics.

The new codes and workflows will constitute a new generation of nine pilot demonstrators (PDs), underpinned by concepts like multi-scale, multi-source, and multi-physics. The PDs will materialize in 15 simulation cases (SCs) that represent capability and capacity use cases of particular relevance in terms of science, social challenges, or urgency. The SCs will produce relevant (European Open Science Cloud) EOSC-compatible datasets; enable services on aspects of geohazards like urgent computing, early warning forecast, and hazard assessment; and foster an emergency access mode in EuroHPC tier-0/tier-1 systems for geohazardous events (earthquakes, tsunamis, and volcanoes). Finally, ChEESE-2P will liaise, align, and synergize with other domain-specific European projects in the solid earth community, including longer-term, mission-like initiatives such as Destination Earth.

HLRS’s role in ChEESE-2P is twofold. Firstly, HLRS is coordinating performance analysis campaigns for the flagship codes. These campaigns will be conducted on EuroHPC systems to reveal performance bottlenecks at scale and to monitor code optimization progress. Secondly, HLRS leads the work package Simulation Ecosystem, which addresses challenges that arise in the scientific simulation workflows within which a main application code is utilized. This work package takes a holistic approach, employing performance analysis and exploiting new methods of simulation setup, deployment, and data analysis to decrease the overall time-to-solution of our workflows.