Glossary¶
This section defines key terms frequently used throughout the REPACSS documentation.
A¶
Account
A project allocation or funding unit used to charge for computing time.
Allocation
The computing time assigned to a project, typically measured in node-hours.
B¶
Batch Job
A job submitted to run without user interaction, scheduled by the system.
C¶
Compute Node
A server node that runs user jobs and provides computational power.
Container
A lightweight, portable software environment used to ensure consistent runtime.
G¶
GPU (Graphics Processing Unit)
A specialized processor optimized for parallel processing, commonly used in AI and simulations.
I¶
Interactive Job
A job that allows users to run commands in real time on compute nodes.
M¶
Module System
Environment management system used to load software dynamically.
P¶
Partition
A subset of nodes within the cluster grouped together for scheduling (e.g., zen4, h100).
S¶
Slurm
The workload manager used by REPACSS to schedule and monitor jobs.
srun
A command to launch parallel tasks under Slurm.
salloc
Allocates resources interactively under Slurm.
sbatch
Submits batch job scripts to the scheduler.