Jobs are managed using slurm, a cluster management system. Generally, jobs are executed via either srun (for directly running e.g. a binary) or sbatch (for running a job script).
See also the sbatch documentation or the srun documentation.
To submit a simple job script, you can use the simple.sh as a template, reproduced below.
Put the following into a bash (denoted by .sh extension) somewhere in your home (~), e.g. ~/experiments/hello-world.sh:
#!/bin/bash
#SBATCH --job-name=hello_world # Give your experiment a name
#SBATCH --mail-type=FAIL # Type of email notification: BEGIN,END,FAIL,ALL,NONE
#SBATCH --mail-user=<YOUR-EMAIL>
#SBATCH --partition=naples,dhabi,rome # Which partitions may your job be scheduled on
#SBATCH --time=1:00:00 # (Optional) time limit in dd:hh:mm:ss format. Make sure to keep an eye on your jobs (using 'squeue -u $(whoami)') anyways.
#SBATCH --mem=1G # Memory limit that slurm allocates
# Memory limit for user program. Must be equal to SBATCH-directive allocation.
# (allows graceful handling of out-of-memory errors in your program.)
let "m=1024*1024*1"
ulimit -v $m
# Print info to stdout on job details
echo "Job-name: $SLURM_JOB_NAME; jobid: $SLURM_JOB_ID; Partition: $SLURM_JOB_PARTITION; No. Nodes: $SLURM_JOB_NUM_NODES; Node: $SLURM_JOB_NODELIST; Memory per Node: ${SLURM_MEM_PER_NODE}MB; Start-time: $(date); Hostname: $(hostname)"
#######################
# Your code goes below #
#######################To schedule one instance of this job to the cluster; sbatch ~/experiments/hello-world.sh. Investigate the resulting output named slurm-$jobid.out.
Now you are ready to begin implementing your own experiments by calling the required programs at the bottom of the hello-world example.
During implementation of your first experiment, refer to the remaining documentation Advanced use, Cheatsheet.