Containing the Future

Container systems, such as those made possible by Linux Containers (LXC) and implemented by the popular Docker software package, make it possible to do reproducible science, data processing for commercial enterprises, visualizations, and everything in between. We will demonstrate work being done in collaboration with the Open Commons Consortium (OCC) to produce a repository of Dockerfiles for commonly used meteorological software. This includes tools such as GrADS, GDAL, MPAS, and AWIPS II. With these containers it becomes possible to offload the burden of getting a specific software package compiled or implemented into a one-time cost that can then be readily reproduced by an entire organization across platforms. We will also showcase how containers are being used by the Environmental Data Commons with orchestration systems such as Kubernetes and Docker Swarm for scalable processing of GOES-16 satellite data. Additionally, we will verify it is possible to run some of the community's existing numerical weather prediction models such as WRF and MPAS in this environment where the resources quickly scale up to meet the demands of the model run and are then destroyed after the run is complete. This flexibility enables cheaper operational and research model simulations without significant maintenance beyond initial configuration.