You are here: Foswiki>Cluster Web>Storage (13 Jan 2022, IngmarCamphausen)Edit

Cluster Storage



Every user has 20GB/25GB soft/hard quota in it's home directory.


Up to now, we restrained from applying rigid disc policies on the cluster storage. To keep this cooperative model going, please use the resources with respect.

CephFS - /data/scratch

On every cluster node you have access to a cluster-wide 330TB CephFS file system with InfiniBand interconnect.
  • scratch Every user has his private directory /home/user/NO_BACKUP - a symlink pointing to /data/scratch/user. Please use /data/scratch for bulk data, job output.

As a rule of thumb, data which is big and can be brought back easily should go into scratch space.

local disc - /data/scratch.local

for temporary use, about 200GB local disc space can be used via /data/scratch.local/ Data found at /data/scratch.local/ and older than a couple of days might be deleted without prior notification.

nfs - /home

  • /home is being mirrored daily to protect your data. Please keep your data usage on /home below 20GB as there is a quota limiting disc usage.

The allegro cluster is designed to be autonomous and mostly independent from the rest of the department's infrastructure. As a consequence, you can not access your nfs home directory from the cluster nodes.

The login node serves as a bridge between your regular $HOME (e.g. /home/mi/user) and your cluster-$HOME. The automount points you know from our linux workstations (e.g. /home, /web, ...) are found by prepending the path with /nfs (e.g. /nfs/home/user).

Bear in mind, however, that the jobs running on compute nodes won't be able to access these file systems.
Topic revision: r2 - 13 Jan 2022, IngmarCamphausen
  • Printable version of this topic (p) Printable version of this topic (p)