Democratizing the computational environmental marine data science: using the High-Performance Cloud-Native Computing for inert transport and diffusion Lagrangian modelling

The use of containers and application containerization is a common approach for cloud-based software architectures due to their flexibility, lightweight nature, simplicity, and scalability. However, a non-negligible feature of today's software applications is the ease of portability between heterogeneous distributed computing clusters, dedicated high-performance computing (HPC) facilities, and public/private computing clouds. This need exists in several contexts, particularly in ocean forecasting. For example, assessing adverse impacts on overall ocean activities requires an accurate prediction of ocean pollution. We propose a new paradigm using a cloud-native approach for cloud-based scalable high-performance computing and define high-performance cloud-native computing (HPCNC). The proposed paradigm's architecture is supported by evaluating the performance of WaComM++ (Water quality Community Model Plus-Plus), a decision-making tool for Lagrangian inert transport and diffusion. This work introduces a new paradigm for computational marine environmental research: High-Performance Cloud-Native Computing (HPCNC). The proposed approach aims to democratize data science, numerical simulation, and AI-based prediction by leveraging cloud and local resources. We show that HPCNC brings tremendous benefits, not only in terms of the performance already achieved by using HPC but also in terms of significant savings in the resources used. The methodology presented here leverages containerization to scale computational resources. The scaling operation is lightweight by implementing a new HPC approach defined as computational power and I/O malleability. Finally, this paper presents our preliminary but encouraging results applying the HPCNC paradigm to the WaComM++ model. WaComM++ is a Lagrangian transport and diffusion model that exhibits high-performance thanks to its hierarchical and heterogeneous parallelization scheme.