Weather Processing in Software Containers across Multiple HPC and Cloud Platforms

Software containers provide the promise of increasing portability of applications by encapsulating the run-time environment in a low overhead, virtualized environment. Not all software container engines, however, are well suited for all applications and platforms. For weather models, mainstream container engines like Docker are not well suited in High Performance Computing (HPC) environments. The development of Singularity from Berkeley Labs attempts to address this issue by designing a container solution geared towards HPC environments. Our research examines the viability of running weather models in Singularity-based software containers in different HPC platforms, including cloud platforms.