Multiscale scientific workflows on high-performance hybrid cloud

Abstract

In this paper, we present an infrastructure for executing multi-scale scientific workflows in hybrid cloud environment. We use an end-to-end modelling of mayonnaise production as an example of an industrially relevant problem involving modelling of materials properties and their corresponding processing methods. We demonstrate the flexibility and modularity of the solution by integrating alternative functional services operating at atomistic, mesoscopic and continuum levels. The services comprise open-source and off-the-shelf commercial software and exhibiting different requirements for parallel computing. To foster reproducibility we employ containers that can be deployed on either public cloud or on-prem resources. We expand on the latter, by discussing a prototype implementation of the on-demand high-performance computing (HPC).