Shared Resources

hpc
supercomputing
An introduction to shared resources in supercomputing, including job scheduling and the Polaris activity webpage.
Author
Affiliation

Sam Foreman

ALCF

Published

July 15, 2025

Modified

August 8, 2025

Supercomputers contain many computer nodes and not every application will use ALL of them. Therefore, we use programs called schedulers that allow users to schedule a job based on how many nodes they need for a specified time.

A job is defined by a user and requires these parameters to be defined:

run-time
How long will this job run? 5 minutes? 60 minutes? 6 hours?
number-of-nodes
How many compute nodes does the application need to run? 5 nodes? 100 nodes? 1000 nodes?

Polaris Activity Webpage

  • We have a page that shows all the current activity on Polaris. The top of the page shows a graphical representation of all nodes. Colors indicate a job running on that node. Below the graphic there is a table that lists running and queued jobs.
    • running refers to jobs running right now on comptuer nodes. If you hover on a colored node in the graphic, it will highlight all the nodes used by the same job AND the job in the running table below.
    • queued jobs are waiting for an opening in which to run.

📊 Live Status of Polaris

Figure 1: See: https://status.alcf.anl.gov/#/polaris

Citation

BibTeX citation:
@online{foreman2025,
  author = {Foreman, Sam},
  title = {Shared {Resources}},
  date = {2025-07-15},
  url = {https://saforem2.github.io/hpc-bootcamp-2025/00-intro-AI-HPC/1-shared-resources/},
  langid = {en}
}
For attribution, please cite this work as:
Foreman, Sam. 2025. “Shared Resources.” July 15, 2025. https://saforem2.github.io/hpc-bootcamp-2025/00-intro-AI-HPC/1-shared-resources/.