Jupyter at NERSC: Background Information¶

Curious about how Jupyter works at NERSC? Want to make more informed decisions about how you use it? If so, read on, this page may help you. But you don't have to read it to make effective use of Jupyter at NERSC.

Jupyter, JupyterLab, JupyterHub at NERSC¶

The terms Jupyter, JupyterLab, and JupyterHub are sometimes used interchangeably, and this leads to confusion. This section is intended to clarify the relationships between these terms an put them in context at NERSC.

Jupyter is a modular, literate computing software ecosystem (programming interfaces, libraries, frameworks, applications) for managing documents called notebooks that contain code, equations, visualizations, interactive widgets, text, and metadata. Notebook files are simple, versatile JSON documents following a well-defined, community-governed notebook format. You can edit and interact with notebooks through applications like Jupyter Notebook or JupyterLab. Also, you can create, manipulate, execute, or transform notebooks non-interactively by using programs that support the notebook format, like nbconvert or papermill. You can share notebooks with others, or convert and publish them in a variety of formats like Markdown, HTML, or even PDF. The Jupyter notebook has become an established, versatile tool supporting reproducible computational science, data science, and education. Around 500-600 users per day, and over 1500 users per month, use Jupyter at NERSC (September 2023).

JupyterLab is a web application that lets you interact with Jupyter notebooks, but it also provides user interface elements for editing text files, using command-line interfaces, browsing the file system, monitoring applications, or visualizing complex data sets. These elements can be arranged and resized as you please, side-by-side, or grouped together using tabs. The goal is to provide a powerful, modular interface for interactive scientific notebook-based workflows. JupyterLab is a popular platform for data science and machine learning. NERSC has supported JupyterLab as its primary user-facing Jupyter application since its release in 2018.

JupyterHub is another web application for spawning, managing, and proxying instances of single-user Jupyter notebook servers, such as JupyterLab, for multiple users. NERSC has operated a JupyterHub instance since 2015. While you need a NERSC account to use Jupyter at NERSC, you can use your NERSC identity or possibly another institution's identity to authenticate yourself to NERSC's JupyterHub at https://jupyter.nersc.gov. From there, you can spawn a JupyterLab server on a login node on a supercomputer, or in a single-node or multi-node batch job. Which kind of server you'll want to start up will depend on your use case, software, and hardware requirements. Be advised that notebook servers running in batch jobs will be charged to a project allocation, most often your default project.