30 likes | 175 Views
Ver 1.0 as of 2 Jul 2014. Reproducibility in Large-Scale Computing: A Dissident View from the Trenches. Doug James Texas Advanced Computing Center djames@tacc.utexas.edu reproducibility@xsede An XSEDE14 Workshop July 14, 2014. Where Do I Stand?*. “No, thank you...”
E N D
Ver1.0 as of 2 Jul 2014 Reproducibility in Large-Scale Computing:A Dissident View from the Trenches Doug James Texas Advanced Computing Center djames@tacc.utexas.edu reproducibility@xsede An XSEDE14 Workshop July 14, 2014
Where Do I Stand?* “No, thank you...” • Run your code • Duplicate your environment • Bitwise reproducibility “Yes, please...” • Implement your algorithm • Portable results • Understand the differences reproducibility > re-enactment *your mileage may vary
What Do I Want to See Next?* • Promote the value of fundamentals a la "Ten Simple Rules" • Especially automation, change management, traceability • No magic bullets imposed from above; let the marketplace decide • Expect clarity • Readable, well-documented source code • Datasets: provenance, formats, meaning of fields • Encourage discussion and reflection • e.g. NSF Grant Proposal Guide ChII.C.2.j (Data Management Plan): "Plans for data management...may include..." [How 'bout a sixth item that says something like...] 6. mechanisms and processes that enhance opportunities for reproducibility." *your mileage may vary