The Pantheon Project: Reproducible Workflows for Extreme Scale Science

Scientific computing is a complex undertaking, involving a myriad of technologies - specialized hardware, software, configurations and version control - all working together to produce large scale data and scientific insight.

An outstanding need is a way to preserve, communicate and validate workflows that cross all of these boundaries, providing both real examples of end-to-end workflows, and permanent records of these workflows for archival purposes.

The Pantheon project aims to meet this need by providing immutable, reproducible, and reusable experiment pipelines on target infrastructures. Panthon provides:

• A community-based standard for reproducible large scale scientific workflows. A workflow in the Pantheon project has been tested against a set of published criteria. If it’s in Pantheon, it works, and is a useful example.
• A federated set of reproducible workflows. Creating large scale scientific workflows is complex, so working examples are a valuable product. We provide these examples as:
  • Examples of real workflows for experimentation, learning and archiving.
  • A permanent working record of milestones.
• Reproducibility for exploratory research. Early software development efforts have unique build requirements and suffer from inconsistent versioning practices. The focus of Pantheon is to capture the state of a working experiment.