Using Library PCs to Crunch Academic Research Data

Jonathan Younker

Brock University

What We're Doing

• Processing research data using high-performance computing (HPC) and many-task computing

• ~120 nodes (468 cores, 632GB RAM)

• Idle time on existing lab PCs, expandable across campus

  • Processing overnight, or during periods of inactivity

  • Student use of PCs always has priority

Why We're Doing It

 

• “A solution looking for a problem”

• Digital scholarship lab initiative

• Integrating research data/methods into library collection

•  

•  

• Becoming ‘baby SharcNet’

  • Accommodating non-STEM researchers

Who's Using It

• Working with Brock University’s Cognitive and Affective Neuroscience Lab to process EEG data

• Prof. Sid Segalowitz & Technician James Desjardins

• http://bit.ly/AccessICA

• Pilot project started spring 2014

• First full dataset processed weekend of September 20th, 2014

How We're Doing It

• Mix of commercial and open source software

  • Microsoft's HPC Pack

  • Matlab/Octave

  • Ability to run Python, C++, etc.

• Node Templates

  • Manual/scheduled/available when idle

• Node Groups (mini clusters)

Challenges

• Steep learning curve

• Proprietary software/licensing

• Existing processes may require change

• Huge support/training/troubleshooting implications

• Need a Mike

To-Do List

• Automation
• Extending to other researchers
• ‘Small’ investment in hardware
• Partnering with Brock ITS/Office of Research Services (via digital scholarship lab initiative)
• Usual SLAs, procedures, etc.

Questions?

• Jonathan Younker (Acting AUL):
  • @jtyounker
  • jyounker [at] brocku.ca
• Mike Tisi (Systems Administrator):
  • @michaelvtisi
  • mtisi [at] brocku.ca