blog

Articles on HPC and cluster workflows

Practical guides, SLURM gotchas, and notes on building tools for computational researchers.

28 Aug 2026

Ask for a GPU slice and queue in minutes

My jobs were not waiting because GPUs were scarce. They were waiting because I kept asking for a whole card to do a fraction of a card's work.

27 Aug 2026

Why generated sbatch scripts go unchecked

Same model, same guess, a completely different failure mode. The argument for generating a small table you can read instead of a long script you cannot.

27 Aug 2026

Stop naming your SLURM job parameters twice

Your script already declares the parameters it reads. Compare that against what your submission supplies and you catch the sweep that quietly used defaults.