Complex-system models describe multiple levels of collective structure and organization. Since such models are usually non-linear and impossible to solve analytically, their analysis is often carried out by simulation on high performance computing environments. Cluster, grid and cloud computing are nowadays adopted, but their use still requires specific technical and methodological skills. That is why high performance computing is not yet part of the daily practices in numerous communities where it would be needed. OpenMOLE (Open MOdeL Experiment: www.openmole.org) is a FOSS framework that implements a solution to this problem. It emphasizes on the easy integration of home-brewed models and on transparent scaling from a single computer up to a grid execution. It makes it possible for the modellers to build workflows for experimenting on their simulation model. OpenMOLE is based on a blackbox approach to embbed models based on very different technologies: java / scala / C / C++ / fortran / scilab / octave / netlogo... Once the user model is embedded in the platform, it can be executed according to various design of experiments, sensitivity analysis, calibration processes... OpenMOLE automatically distributes the numerous executions of the model on a distributed environment specified by the user, such as a cluster or even on the 100,000+ cores available on the European computing grid. The platform takes care of software installation, files transfert, job failures, rendering the distributed execution entierly transparent.
Speakers: Romain Reuillon Mathieu Leclaire