Towards Choreographies for Multi-Scale and Multi-Physics Simulations

In previous work in the context of SimTech an approach for modeling and adaptation of single simulations implemented as scientific workflows has been proposed [SK13]. However, there is no user-friendly tool support for modeling and execution of multi-scale and multi-physics simulations. Coupling of different simulation experiments is done manually. This is cumbersome and error-prone. We aim to provide scientists with means to model multi-scale and multi-physics simulation experiments as choreographies. Chore-ographies are a concept known from the business domain that en-ables independent organizations to collaborate and reach a com-mon business goal, e.g. as manufacturers and suppliers in a supply chain. Therefore, the main goals of our work are: Provide a meta-model and language for simulation choreogra-phies Investigate how the coupling between different simulations can be automated with simulation choreographies Investigate how uncertainties (e.g. regarding data quality, re-source availability) can be modeled with choreographies Investigate how multi-scale and multi-physics simulation chore-ographies can be adapted during runtime and provide neces-sary modeling constructs Provide a user-friendly modeling tool that implements the re-searched concepts State of Current Work We have identified two modeling approaches and implemented a first version of a choreography modeling tool. Fig. 1 shows two possible approaches for modeling multi-scale and multi-physics sim-ulations. Modeling can be conducted in a top-down or bottom-up style. Scientific multi-scale and multi-physics problem Manual modeling with choreography editor Automatic transformation Manual refinement Modeled choreography Abstract workflows Refined & executable workflows Scientific multi-scale and multi-physics problem Automatic transformation Derived choreography Executable workflows a b models Legend Communication activity Simulation step activity Control flo