Lazzaro et al. unveil portable software ecosystem for reproducible HPC workflows

Scientists regularly juggle laptops, workstations, supercomputers, and cloud resources, often needing to adapt their code for each environment. A new paper by Alfio Lazzaro, Utz-Uwe Haus, Sandrine Charousset, and Nina Mujkanovic presents a software ecosystem designed to eliminate this friction. Published on arXiv (2607.18929), the system offers a modular, unified command-line interface that allows users to interact with their workflows consistently across all hardware platforms. The key advantage is portability: researchers can develop on a local machine and deploy on a high-performance cluster (HPC) or cloud with zero changes.

Tested as part of the plan4res EU H2020 project, the ecosystem has proven capable of handling real-world computational workflows. It builds on containerization and reproducibility best practices, ensuring that results are verifiable regardless of the underlying hardware. The authors emphasize that the approach is extensible to other projects with similar multi-platform requirements. By abstracting away hardware-specific details, the software reduces debugging time and accelerates the path from prototype to production. This is particularly valuable for distributed and parallel computing (cs.DC), where heterogeneity is the norm. The paper includes four figures and is linked to a Springer Nature proceedings chapter from ISC High Performance 2024 workshops.