Enabling Interactive Notebooks on Supercomputers with Jupyterhub

Interactive notebook systems such as Jupyter represent a new paradigm in web science gateways that can combine interactive code execution with data analysis and exploration. In our work we demonstrate how one can create and manage interactive notebooks in a multi-user supercomputing environment using the Jupyterhub platform. We describe our architecture along with custom modules that we developed for Jupyterhub to manage authentication, notebook execution and interaction with the job queueing system. We illustrate the power of this system through the OpenMSI use case, and outline future directions for this effort.