Essentially IBM’s GPFS (re-branded to “Spectrum Scale”) was MOGON I’s file system, while MOGON II uses Lustre file system. Both are parallel file systems, meaning that the data is accessible on every client (indicated by the small boxes in the figure).
In either case:
Our filesystems are not a wastebasket for your data. The ZDV offers an archiving system – it is your responsibility to occasionally clean up or archive data.
HOME directory on MOGON for each user is a link to
/gpfs/fs1/home/<username> visible as
Purpose and Characteristics
Whilst the project directories (resp. their underlying file systems) are optimized for I/O of many big files, the home directory obviously cannot match this property: Small scripts, configuration files, aggregated results, etc. – this is what a home directory is for. Put negatively, this requires two restrictions or policies:
The home directory is not for heavy I/O , particularly not on MOGON II, as it is exposed via NFS to this cluster.
Looking on it the other way around, scripts, code, config files, small text files may go here, without second thoughts.
Currently the home directories have the following default setting:
- One 150.000 files per user
- a quota of $50 GiB$.
Project Directories on MOGON II
The project directory on MOGON II currently resides under
The purpose of this filesystem (and the related mountpoint) obviously is to place all data relevant to the project, which is not volatile (in contrast to the scratch filesystem, see below).
See our quota page for further information.