On the efficient use of HPC ressources: Disaggregation of the full waveform inversion algorithm
Résumé
This study introduces a novel approach to enhance the efficiency of the full waveform inversion (FWI) algorithm in high-performance computing (HPC) clusters. By disaggregating the FWI scheme into three stages and implementing a workflow manager, the methodology offers flexibility in task distribution and improves fault tolerance. The granularity, adjustable by the user, enables efficient subdivision of large jobs, facilitating backfilling by the batch scheduler. Despite increased disk input/output, potential advantages in fault tolerance and system workload are highlighted. The workflow manager oversees job submission, ensuring correct execution and data organization. Overall, the methodology optimizes HPC infrastructure, addressing underutilization issues and contributing to more efficient seismic data processing, particularly in complex subsurface imaging applications.
Origine | Fichiers produits par l'(les) auteur(s) |
---|