Workflow

The following flowchart demonstrates the design of AMEP. The three layers “Data Storage”, “Data Analysis”, and “Result Storage” outline the unification of storing and analyzing simulation data and analysis results. The simulation data is stored in HDF5 files together with various metadata such as simulation parameters, simulation scripts, software information, and author information. For the data analysis, this data is loaded into Python as a Trajectory object which is then either used as an input to the evaluate module or to directly access individual frames of the trajectory and the contained data as NumPy arrays, which can then be analyzed using AMEP’s various submodules. Analysis results are stored in an HDF5-based data format as well. This format can be used as a container for multiple analysis results and can be loaded with AMEP as a DataBase object, from which individual observables can be returned as EvalData object. The latter again allows to return the data as NumPy arrays for further processing and visualization. The red arrows show a typical workflow of first loading the trajectory, second, calculating an observable, third, storing the result in an HDF5 file, and finally visualizing the results.