Overview
The main objective of mUQSA is to provide an easy-to-use platform for efficient performing commonly requested UQ and SA studies of diverse application types. To this end, mUQSA is divided into two main layers: User Interface (UI) layer, and the Execution layer.
The UI layer has two main duties:
- definition of all input parameters and configuration settings for UQ/SA scenario,
- presentation of results for further analysis, once the required calculations have been completed.
In turn, the Execution layer’s role is to efficiently execute required computations in accordance with the selected method and user-defined input data.
Given this context, an overall picture of the execution flow in mUQSA can be visualised as in the figure below:
The Scenario Definition and Analysis steps correspond directly to the two above duties of UI. Scenario Definition’s aim is to collect all specific data about the scenario from a user, including selection of a Method and definition of Parameters and then to push it for fully automatic processing on an HPC machine. Analogically, the Analysis step receives results from the processing on HPC machine and displays an interactive web page with condensed information for further analysis by a user.
The rest of steps of the workflow has a strictly computational character and belongs to the Execution layer. Sampler generates sets of inputs for model evaluations in accordance with a selected method. Then, as many streams composed of Encoder, Execution and Decoder as a number of required evaluations is launched, likely in a parallel way. Next, the output data from the evaluations is merged together by Collator and analysed programmatically by Analyser, again in accordance with the selected method. Finally, the results are transformed into the form attractive to humans by the Generator.