Architecture and deployment details of scalable Jupyter environment at Kurchatov Institute supercomputing centre

Abstract

Jupyter notebook is a popular framework for interactive application development and data analysis. Deployment of JupyterHub on a supercomputer infrastructure would allow to combine high computing power and large storage capacity with convenience and ease of use for end users. In this work we present the architecture and deployment details of Jupyter framework in Kurchatov Institute computing infrastructure. In our setup we combined JupyterHub with CEPHfs storage system, FreeIPA user management system, customized CUDA-compatible image with worker applications and used Kubernetes as a component orchestrator.