The Pilot Lab Exascale Earth System Modelling

The Pilot Lab Exascale Earth System Modelling (PL-ExaESM) is a “Helmholtz-Incubator Information & Data Science” project and explores specific concepts to enable exascale readiness of Earth System models and associated work flows in Earth System science. PL-ExaESM provides a new platform for scientists of the Helmholtz Association to develop scientific and technological concepts for future generation Earth System models and data analysis systems. Even though extreme events can lead to disruptive changes in society and the environment, current generation models have limited skills particularly with respect to the simulation of these events. Reliable quantification of extreme events requires models with unprecedentedly high resolution and timely analysis of huge volumes of observational and simulation data, which drastically increase the demand on computing power as well as data storage and analysis capacities. At the same time, the unprecedented complexity and heterogeneity of exascale systems, will require new software paradigms for next generation Earth System models as well as fundamentally new concepts for the integration of models and data. Specifically, novel solutions for the parallelisation and scheduling of model components, the handling and staging of huge data volumes and a seamless integration of information management strategies throughout the entire process-value chain from global Earth System simulations to local scale impact models are being developed in PL-ExaESM. The potential of machine learning to optimize these tasks is investigated. At the end of the project, several program libraries and workflows will be available, which provide the basis for the development of next generation Earth System models.

In the PL-ExaESM, scientists from 9 Helmholtz institutions work together to address 5 specific problems of exascale Earth system modelling:

• Scalability: models are being ported to next-generation GPU processor technology and the codes are modularized so that computer scientists can better help to optimize the models on new hardware.
• Load balancing: asynchronous workflows are being developed to allow for more efficient orchestration of the increasing model output while preserving the necessary flexibility to control the simulation output according to the scientific needs.
• Data staging: new emerging dense memory technologies allow new ways of optimizing I/O operations of data-intensive applications running on HPC clusters and future Exascale systems.
• System design: the results of dedicated performance tests of Earth system models and Earth system data workflows are analysed in light of potential improvements of the future exascale supercomputer system design.
• Machine learning: modern machine learning approaches are tested for their suitability to replace computationally expensive model calculations and speed up the model simulations or make better use of available observation data.