Verbundprojekt: Auto-Tuning System für eine effiziente Informationsschnittstelle für Exascale Supercomputer - TOPIO

Abstract

As available computing power continues to increase, improvements in weather forecasting have led to a dramatic increase in data generated by each simulation. However, the I/O performance and storage capacity of modern HPC systems have not kept pace with this rapid expansion of data. Therefore, optimized I/O configurations and sensible data reduction become necessary to improve overall computational resource utilization. In this context, the work at the University of Hohenheim in collaboration with HLRS, focuses on:

1.1. HPC I/O Optimization: Improvement of HPC I/O performance to reduce the overall runtime of the Numerical Weather prediction model MPAS. 1.2. Data Compression: Compression of the large output data generated by the weather model to reduce storage requirements.