Webinar – Beyond Speed: Integrating Energy Considerations into I/O Performance Analysis for SKA Codes

Beyond Speed: Integrating Energy Considerations into I/O Performance Analysis for SKA Codes

Astronomical data processing codes for the analysis of massive observations are inherently demanding in terms of computational resources and input/output (I/O) operations. The constant increase in data volume and complexity is creating significant bottlenecks, affecting not only the execution time of applications on supercomputers, but also their energy consumption, which is a crucial issue for the sustainability and efficiency of HPC infrastructures.
This presentation will introduce a methodology for characterising the impact of I/O on overall computation time and, in an innovative manner, on the energy consumption of compute nodes and parallel storage systems such as Lustre.
Through concrete case studies drawn from SKA codes, we will show how identifying data access patterns and using advanced profiling tools can reveal opportunities for optimisation. The aim is to provide developers and users of astronomical codes with a roadmap for designing more efficient applications, reducing both the time-to-solution and the energy footprint of their research.