Slurm is a job scheduler which allows you to access compute nodes on the cluster.
There are three partitions:
The dev partition contains four 8-core nodes for interactive work.
When you submit jobs to the low priority queue, e.g.
sbatch -p lowprio --array=1-16 ./my_not_important_job
if there are free nodes, the jobs start.
When someone submits job another job
srun -p dev ./my_important_job
the running low-priority job is requeued to make space for the important job.