Siboka: a next generation scientific computing workflow toolkit at Lawrence Livermore Laboratory

The Department of Energy research laboratories are moving rapidly towards exascale supercomputing to help solve large and difficult problems in both national security and fundamental science.  To harness the full power of these extraordinary machines, new tools must be developed to increase user efficiency and enable greater reproducibility of results. The Siboka project (from the Abenaki word ‘Sibokan’, meaning to travel/work a river) focuses on these key issues, enabling reproducible, pedigreed computational science through advanced analysis and machine learning capabilities in service of the LLNL mission space and the user community.  We will describe the common Application Programming Interfaces we are developing for storing and querying simulation data by providing federated database capabilities for scientific workflows that leverage SQL and NoSQL data stores.  We will overview the development of the following capabilities: - Mnoda (Abenaki for ‘basket’), a common JSON schema for capturing and maintaining simulation information intended to facilitate identification of "interesting" simulation runs.   - Sina (from the Abenaki ‘Kina’, meaning to-see), Simulation INsight and Analysis is a Python package for managing data stored using the Mnoda schema. We will also describe how we are using Jupyter Notebooks to serve as our primary vehicle for providing examples and tutorials on performing Sina queries, visualizing the results, and interacting with the data.