The mésocentre provides two temporary storage spaces dedicated to computing. They are accessible from the connection front-ends and to the skylake, dev, smp-opa, kepler, pascal, volta and visu partitions.
Copying files between these spaces or between one of them and /home is possible.
This partition offers good disk read/write performance and is suitable for all types of computation.
A 6TB quota is automatically allocated to users when their mésocentre account is created.
To use it, no special instructions are necessary.
This is a temporary storage space and the mésocentre does not guarantee the sustainability of the data.
The data stored in this space are deleted when the user account is closed.
For reasons of system performance, users are requested to keep only regularly requested data.
This partition offers very good disk read/write performance and is intended for applications performing numerous I/O operations on files.
Using the partition:
- When submitting the job, the user must specify the disk space required by the following command #SBATCH -L scratchfast: number_of_Go (ex #SBATCH -L scratchfast: 10 to request 10GB). If the requested disk space is not available, the job is queued. The output of the squeue command then indicates Licenses in the NODELIST / REASON column.
As this space is shared, requests for space must remain reasonable so that all users can benefit from it.
- When the job starts, a temporary directory /scratchfast/$SLURM_JOB_USER/$SLURM_JOB_ID/ is created for the duration of the job.
- At the end of the job, this directory is automatically deleted along with all the data it contains. The user must therefore make a copy of the data to be kept before the job is complete. If not, or if the job ends before the copy command is run (for example because of a too short walltime), the data in the temporary directory are lost.