Comparing Open Ondemand and Jupyterhub as Interactive HPC Gateways

Michael Milligan, Graham T. Allan, Nicholas J.H. Dunn, Matt Mix

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The Minnesota Supercomputing Institute (MSI) at the University of Minnesota has adopted a goal of supporting Interactive high performance computing (HPC) as a first class service. For several years MSI has used Jupyterhub to provide a web gateway to interactive Jupyter notebook environments on HPC resources. In the past year MSI has additionally adopted Open OnDemand as a general purpose web gateway for interactive HPC services. This places MSI in the position to provide an experience-informed comparison of two popular open-source platforms for interactive web-based access to HPC resources. Open OnDemand is designed specifically to provide access to HPC resources, and includes facilities to manage files and jobs, access in-browser command line sessions, and provide virtual desktop access to GUI applications as well as proxied access to web applications such as Jupyter servers. Jupyterhub, in contrast, is designed to provide multi-user access to Jupyter servers, and when configured with the Batchspawner extension is able to manage such servers on scheduled HPC cluster resources. Jupyterlab, when accessed through a Jupyterhub so configured, likewise offers access to in-browser facilities such as command line terminals and file management, and can proxy access to some other web applications, in addition to supporting the main Jupyter Notebook application. From a systems perspective, both platforms provide a highly configurable base on which to build customized gateway solutions, but the differing focus of the two projects results in different characteristics and challenges. Despite different origins and project focus, Jupyterhub and Open OnDemand have evolved to be able to provide strongly overlapping services to users of an HPC center. This poster will compare the characteristics of the two platforms and the different advantages and limitations of each, from the perspective of an HPC center that has deployed both into production for use by a large base of research users.

Original languageEnglish (US)
Title of host publicationPEARC 2023 - Computing for the common good
Subtitle of host publicationPractice and Experience in Advanced Research Computing
PublisherAssociation for Computing Machinery, Inc
Pages466-468
Number of pages3
ISBN (Electronic)9781450399852
DOIs
StatePublished - Jul 23 2023
Event2023 Practice and Experience in Advanced Research Computing, PEARC 2023 - Portland, United States
Duration: Jul 23 2023Jul 27 2023

Publication series

NamePractice and Experience in Advanced Research Computing

Conference

Conference2023 Practice and Experience in Advanced Research Computing, PEARC 2023
Country/TerritoryUnited States
CityPortland
Period7/23/237/27/23

Bibliographical note

Publisher Copyright:
© 2023 Owner/Author.

Keywords

  • HPC
  • Jupyter
  • Jupyterhub
  • Open OnDemand
  • interactive computing
  • science gateways

Fingerprint

Dive into the research topics of 'Comparing Open Ondemand and Jupyterhub as Interactive HPC Gateways'. Together they form a unique fingerprint.

Cite this