Dynamic Workflows for Multiphysics Design

International audienceLarge-scale multiphysics applications, e.g., aircraft flight simulation, require several layers for their effective implementation.The first layer includes the design of efficient methods for mathematical models, numeric problem solutions and data processing. This involves the optimization of complex application codes, that might include intricate and distributed execution tools. This includes the asynchronous execution of coordinated tasks executing in parallel on remotely connected environments, e.g., using grid middleware. The third layer includes sophisticated tools that allow the users to interact dynamically in explicit and coordinated ways to design new artefacts, e.g., workflow systems. This presentation is devoted to the third layer where sophisticated application codes are deployed and must be run cooperatively in heterogeneous, distributed and parallel environments. It is assumed here that the first and second layers are implemented and running to support the execution of the workflows. The paper focuses on some open challenges in deploying and running distributed workflows for mutiphysics design: resiliency, exception handling, dynamic user interactions and high-performance computing