Patrick Chabrier (séminaire interne Record)
Résumé : Like in the REDELAC project whose objective is to assess the impact of the future climate and to anticipate the adaptive evolution of dairy farms in the Pays de Fougères, located in the Brittany region of France, virtual scientific experiments are involved and they rely on the collaborative development of workflows of simulations on appropriate computing resources. The first stage of the REDELAC project is based on the use of the Stics simulator, because of its ability to represent a wide variety of rotational cropping systems. Moreover, the size of the overall simulation plan is estimated at around ten million simulations, making it very difficult to implement on an individual workstation. In this working context, the first need is for a collaborative working environment between Stics and SticsRPacks development engineers, the scientists in charge of the REDELAC project and the engineers in charge of simulation workflow development and execution. The second priority is to benefit from solutions enabling the deployment and development of heterogeneous components in the form of processing workflows on high-performance HPC resources. And finally, in the context of a scientific project, mobilizing simulators and tools developed in a scientific context, we need to be able to quickly take into account inevitable evolutions in order to redeploy any part of the workflow, and if necessary to replay the various processes sparingly, as long as the project is not finished. This means that reuse during the project, but also reuse for communication and enhancement purposes, is an important issue. To meet the requirements of the REDELAC project we are using a framework based on Github and Gitlab software forges, Docker and singularity Containers, CI/CD pipelines, the meso@LR HPC cluster and the SIWAA Galaxy Web workflow System. In this communication we explain how the design of this overall framework sustain the scientific project in many ways.