Fig. 3
Minimal example. Illustration of the Archivist’s functionality in a simple example use case. In a parameter scanning experiment, several instances of a model with different configurations (parameters) are simulated (“Simulation 1”, …, “Simulation N”; blue boxes on the left). During each simulation, configuration and performance information are recorded and stored in a (raw) metadata archive (yellow). After each simulation, the stored metadata is post-processed (“Metadata post-processing 1”, …, “Metadata post-processing N”; red): first, the relevant information is extracted by user-defined Parser classes (gray box parsers.py). Non-relevant information is discarded (see light gray text in the raw metadata files). The extracted metadata are then structured according to a provided data template (gray box data_template.json). Finally, the simulation results are annotated with the structured metadata and stored in a database (red cylinder). After all simulation and metadata post-processing instances are finished and their corresponding results are stored in the database, the annotated data can be queried and presented (green).
