Speaker
Description
We present a JupyterHub platform that was built for the Direção-Geral do Território (DGT) of Portugal, offering accessible interactive geospatial analytics and HPC within one browser-based ecosystem. The platform uses pre-configured Python and R environments allowing users to discover, process, and visualize their geospatial data using STAC services, Sentinel data access utilities, and an S3 object storage solution. It allows to perform workflow operations in satellite images, LiDAR, as well as other raster and vector data. Reproducible and persistent workspaces created with containerized software allow users to seamlessly work with these data. The architecture relies on locally launched Docker-based notebooks and HPC sessions running under the control of Slurm inside the Apptainer (Singularity) containers. User authentication via Keycloak and predefined resource profiles is provided by a unified web interface, as well as automated job submission, facilitating HPC resource management. Future development is targeted at Kubernetes deployment of the platform with improved orchestration capabilities, scalability, and fault tolerance while preserving Slurm HPC functionality.