Previous work has introduced a Cloud platform model and budget-aware static algorithms to schedule stochastic workflows on such platforms. In this paper, we compare the performance of these algorithms obtained via simulation and via execution on an actual platform, Grid'5000. We focus on DIET, a widely used workflow engine, and detail the extensions that were implemented to conduct the comparison. We also detail additional code that we made available in order to automate and to ease the reproducibility of such experiments.