disk_areas.md

Abel Disk areas that are relevant for your work at the Veterinary Institute

Summary :

directory	path	what	characteristics
$HOME	usit/abel/u1/username	"home-home" area	500 GB - login area - NOT FOR Data Processing - backed-up
work	/work/projects/nn9305k	common resources (eg. softwares, databases)	10 TB - faster disc, closer to nodes connection
project	/projects/nn9305k	Project data area (backed-up, raw data, results)	30 TB - slower disc
project-home	projects/nn9305k/home/username	Non-collaborative projects	subdirectory of project area
$SCRATCH	/work/jobs/JOBID.d	temporary area on computing nodes: FOR PROCESSING DATA = RUNNING JOBS	fastest
$USERWORK	/work/users/username	sharing data among different jobs = staging	do not backed-up - removed after 45 days

Storage Areas

1. $HOME: "home-home" directory (on login nodes): 500GB - backed up regularly: slow and timewise costly.

• If you need to store files that do not require backup: put them in a nobackup directory
• Recommended for: software, job configurations, job preparation, debugging ...
• All Veterinary Institute data you work on has to be accessible to the Veterinary Institute user-group. So you can’t store them in your "home-home".

2. work area: /work/projects/nn9305k will be used to contain:

• Softwares and libraries
• Public reference data, template files, adapters files ...

3. project area: /projects/nn9305k will be used to contain:

• Project data, including raw sequencing files (zipped/tared:compressed and decompressed)
• Data analyses that have been performed
• ... this is where the work will NOW be done.

4. project-home area: projects/nn9305k/home/username: sub-directory of the Veterinary Institute project area. Here should all data from projects that do not require collaboration with other users go. NB: Data from collaborative projects must be placed in a directory directly under project area.

NB: There is also a version of NCBI databases hosted on Abel: at /work/databases/bio/ncbi

---

Computing Areas (computing nodes)

5. $SCRATCH: when a job is started, a temporary area on $SCRATCH is automatically created. It is also automatically deleted when the job finishes. This allows the jobs to run faster without interfering with the work of others. You can access $SCRATCH for status monitoring of running jobs: with /work/jobs/JOBID.d where JOBID.d is the ID that has been provided when you submitted the job in the queue system SLURM (thus accessible only during job's lifetime).

6. $USERWORK If files are needed for more than one job, they are staged in here. Bifrost is configured to set files in $USERWORK by itself, BUT using some other softwares/pipelines might requires that you copy needed files here. You should run analysis from this area (via SLRUM script or other mechanisms). Path: /work/users/username OR type $USERWORK instead of path.

• NB: Abel managing system automatically delete files older than 45 days (those files are not backed-up).

For more details you can look at: Managing Data on Abel and more generally at Abel User Guide. You can also look at Computer Resources at CEES.

SLURM is the queue system and scheduler daemon (program/process which always runs in the background on Abel). It allows optimizing computing resources and schedule when jobs are to be run on $SCRATCH.- All long jobs should be sumbited to to the queue system SLURM (and are run on $SCRATCH).

• There are 2 ways to use slurm SLURM queue system: 1. interactively with the command: qlogin and 2. by submitting a script: using sbatch.

• when you use qlogin you actually are logged on a computing node that now executes commands from the shell. Usage: small tests (testing a working process OR using interactive programs).

• when you submit a script to SLURM, it will be scheduled to run on a computing node when resources are availables. Usage: heavy jobs.

• NB: Small tests outside SLURM are tolerated (processes taking more than 30 min are automatically killed)