The beginnings of SCOPED: Seismic COmputational Platform for Empowering Discovery
Abstract
Seismology is a data and model-intensive field. Data-driven computations benefit from Cloud Computing because observational seismology relies on horizontal scalability. Model-driven computation benefits from High-Performance Computing because simulations necessitate large memory and a high number of workers for efficient parallelization. Our project SCOPED (Seismic COmputational Platform for Empowering Discovery) aims to bridge both observational and theoretical fields by building a CyberInfrastructure (software, data) as a service to the seismic community. The challenges ahead are the benchmarking of open-source codes in cloud and HPC environments, the containerization of these codes adapted to various computing architectures, and the choice of data formats adapted to both Cloud and HPC computing, among others. This presentation will cover the beginnings of the project in three main directions. First, we present the successes and failures in the implementation and scalability of the different full-waveform inversion workflows across multiple platforms. Second, we briefly discuss opportunities to use the ASDF container, as ready from a cloud-optimized H5 library (H5coro) as a standard format for ambient seismic noise analysis, seismic event location, full-waveform modeling, and machine learning applications both in the Cloud and on HPC clusters. Third, we present cloud-based workflows for data streaming and processing in both Julia and in Python (MsPASS).
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.S15E0296D