Fine-tuning of a weather model requires immense computational resources, however, such capacities are usually available on non-homogeneous IT platforms. In addition, development and operational application are typically performed on different, heterogeneous systems (from laptops to dedicated HPC servers or cloud computing environments). To manage scalability and platform independent portability, a new layer – supporting state-of-the-art software container technology and batch processing – has been introduced. Encouraged by prior successful benchmark tests of the WRF model, the effect of model setup has been investigated over 10 different cases, tested on 30 different configurations. Including different parameterizations, the results of 300 different runs can be compared in a uniform database, yielding a sufficiently wide pool of samples in order to obtain the configuration of the modeling system optimal to the scope of our research, based on a relatively objective selection method. Continuously expanding database of near real-time preliminary outputs gives the opportunity for run-time steering of the experiments. This research currently benefits the development of an aviation meteorological support system, in the meanwhile, our contributions could be applied in an even wider aspect, either from the applicability of big data technology point of view, or with respect to the given best practice model setup.