Introducing HPC to new PhD students

Two challenges PhD students encounter are:

• Knowing what up-skilling to do when you start
• Not being aware of better ways of doing things or feeling you have the time to up-skill once you’ve gotten very busy

To help address the above for a particular case I spoke with a new cohort of PhD students today (from the Speech and Language Technologies Centre for Doctoral Training) to explain what high-performance computing (HPC) is and why they might care. The hope is that they will now be able to include HPC in their training plans once they recognise problems that HPC might be well-suited to helping with.

The talk covered:

• Pros and cons of HPC vs personal laptops/desktops
• Local HPC resources
• General HPC architectures
• Jupyter as an alternative interface to HPC
• Transferring and storing data
• Running jobs:
  • Requesting resources
  • Batch jobs vs interactive use
  • Job management
• Estimating required resources
• Accessing/installing research software
• Optimisation/parallelisation (various options)
• Other HPC systems/environments with more resources
• Getting help
• Local community resources (e.g. mailing lists, talks, workshops)
• Info about the RSE team.

The slides include lots of detail I didn’t cover ‘in person’ but told the students they could explore if interested/a need arises.