Skip to content

Cluster data storage

GPFS file allocations

The main storage for the VACC cluster is housed on GPFS file servers, and organized into two main storage areas: home space (/gpfs1) and scratch space (/gpfs2). Usage quotas are imposed on both 'filesystems'; The quota limits for both can be found at the VACC cost and payment page.

Home space

All user home directories are stored on gpfs1. For the user grundoon, the actual home directory will be /gpfs1/home/g/r/grundoon. The system will report the home directory as /users/g/r/grundoon, where /users is a symbolic link to /gpfs1/home. Home directories are created with permissions that enable all users in the same PI group to read their contents.

In addition to the home directories for users, each PI account has a shared data directory. Again, if the PI's NetID is grundoon, then the shared directory is at /gpfs1/pi/grundoon. Shared directories are created with permissions that enable any member of the PI group to read and write to them.

Use shared directories!

The shared directories are highly recommended for any data that should remain with the group when individual users leave. For example, any data generated by external instruments such as sequencers or MRI scanners should be put into shared directories.

All home space, including the PI's shared data directory, is backed up, so some protection against loss is included.

Quotas apply to the amount of space occupied on /gpfs1, and to the number of files stored on /gpfs1, that are owned by the group in aggregate. When calculating the space occupied, or the number of files, GPFS does so for all files owned by the group regardless of location on the server.

Calculations of space and file count are done independently for each file system, e.g., /gpfs1 is independent of /gpfs2.

Scratch space

Many analytic processes will need more space than can be accommodated on the home file servers, so we also provide 'scratch' space on a different file server. The scratch space is temporary and should not be used for long-term storage. The VACC reserves the right to delete any or all data from scratch file systems without notice, if needed.

The data on the scratch file servers is not backed up.

As with home, every user on the VACC cluster will have their own scratch space. When an account is set up, a symbolic link is created in the home directory pointing to the user's scratch directory. This enables easy access using ~/scratch or $HOME/scratch. The actual location is /gpfs2/scratch/<netid>.

As with the home space, there is also a shared scratch space. Using our PI, grundoon, it would be at /gpfs2/scratch/pi/grundoon and would be writable by all members of the PI's group.