The FhGFS file system is a parallel file system especially designed for high performance applications. The User and the applications see one single file system that is provided by a number of fileservers. Depending on the configuration, the data is distributed over several file servers which improves read/write performance significantly.

The FhGFS can be used as a global scratch directory. It is accessible from all login nodes as well as from all compute nodes. User jobs with large requirements in temporary data can hence be copied to the FhGFS file system and be submitted from there.

Currently FhGFS has no quota support, so one user could theoretically fill up the whole file system. So please be care full when producing a lot (several tens of TBs) of data. Nevertheless, we have created a folder for each user which has the same permissions as the $HOME folder (owned by the user, project group has reading permissions). It is stored in the environment variable $GLOBAL and points to /global/lv<project_id>/<username>.

currently the FhGFS have the following sizes:

VSC-1: 30TB on 2 Fileservers

VSC-2: 99TB on 6 Fileservers