Title:
Addressing Barriers to In-situ Use of Jupyter Notebooks in HPC Simulations

Description:
Jupyter Notebooks offer users a reproducible, interactive environment and are heavily used in the big data world as a front-end for Spark and other parallel data analysis tools. At LLNL, we see tremendous potential in Jupyter as a front-end for in-situ analysis of large-scale parallel simulation codes. HPC centers are notoriously complex and to address deployment concerns we extended JupyterHub to provision certificates for authentication and developed a thin kernel client strategy, which allows a notebook to connect to a running simulation (or other backend); this approach has been used successfully with an existing LLNL code.



Presenter(s):
 Speaker: Thomas Stitt, Lawrence Livermore National Laboratory
 Speaker: Thomas Mendoza, Lawrence Livermore National Laboratory
 Speaker: Cyrus Harrison, Lawrence Livermore National Laboratory
 Speaker: Todd Gamblin, Lawrence Livermore National Laboratory