Best Practices#

Note

Disclaimer It’s very difficult to specify exact best practices that cover every scenario. Ask the other people on your team, your local admin, team lead, etc.

Start Small and Save Money#

Everyone likes saving money!

Typically, I login, and choose a smaller, a t3a or m5 instance with 16GB or 32GB of memory. That is my workstation. It’s like when I had a job and I would walk into the office and turn on my computer.

Submit Larger jobs to the Scheduler#

Any large scale jobs I will submit to the SLURM scheduler using sbatch. At an absolute minimum you need an executable shell script.

#!/usr/bin/env bash

date
sleep 10
hostname
date

echo "goodbye"

Save this text to test-sbatch.sh and run chmod 777 test-sbatch.sh. Then submit it with sbatch ./test-sbatch.sh. That’s it!

Exceptions#

Of course, sometimes people need to do interactive analysis, and they need a larger instance. Just don’t choose a larger instance unless you have a need for it.