February 4, 2019 - by CSCS
Even by the most conservative assumptions, computers capable of performing 1 ExaFlop/s (10^18 floating point operations per second) will be on-line in the next five years, potentially allowing scientists to solve hitherto unsolvable problems. But a central concern remains about whether individual scientific domains will be able to make good use of this exceptional performance. The Climate and Weather Prediction community is preparing for exascale now through the EU-funded ESiWACE-2 project. Within this project, ETH Zurich (CSCS) and MeteoSwiss will collaborate with partner institutions throughout Europe — e.g., the German Climate Computing Center (DKRZ), Centre National de la Recherche Scientifique (CNRS-IPSL), the European Centre for Medium-Range Weather Forecasts (ECMWF), and the Barcelona Supercomputing Center (BSC) — to prepare atmospheric and oceanographic models for the coming generation of computers. To this end, current models will be ported to pre-exascale machines, such as Piz Daint, and new software tools and middleware will be investigated to enable the transition to the emerging exascale architectures, e.g., container technology.
Containers allow operating-system-level virtualization without the heavy overhead of virtual machines which offer the full emulation of a computer system. CSCS will help modeling teams from the project partners to "containerize" their models to allow them to run seamlessly on a variety of platforms.
As part of the activities around this effort, a container unit for the ESiWACE-2 summer school is being planned, as well as a one-week container "hackathon" to make quick progress on applications by teaming container experts with the modeling teams. Finally, CSCS will also perform one-on-one support of the modeling teams to help them achieve their milestones.
For more information on the project please visit ESiWACE website.