School of Informatics and Computing

The following disk storage space is available to users of the RI clusters:

• Home Directory - Everyone on the clusters has a "home" directory, which is your current working directory when you login. This is hosted on a single NFS file server. If you do disk intensive operations from many nodes using your home directory you will over-load the NFS server, negatively impact all cluster users, and get very poor results. For more I/O intensive use, you are encouraged to use PVFS. Please note that your RI cluster home directory is NOT BACKED UP.

• PVFS - All users have a directory named /pvfs/USERNAME. This uses the the Parallel Virtual File System (PVFS) to distribute the I/O load across multiple servers. The /pvfs filesystem is currently accessible from all nodes in the odin cluster. There are three PVFS I/O nodes (odinnfs2, odinnfs3, and odinnfs4) which also act as meta-data servers and this provides a combined total of 22TB of storage capacity. This filesystem is NOT BACKED UP.

• Data Capacitor/Lustre - The Data Capacitor is accessible using the path /N/dc on the tyr and sif clusters only. This filesystem is NOT BACKED UP.

• NFS - In general, the NFS filesystems on the other CS systems (such as your CS home directory) are not accessible from the RI cluster systems. The notable exceptions to this are the head nodes odin and sif (but none of the compute nodes), jord, and idun. On these systems, you can access any NFS-exported filesystem using the path /nfs/SERVER/PATH, where you replace SERVER with name of the file server and PATH with the path to the directory on that server. Here are the most common paths you are likely to want to access:

  CS Home Directory	/nfs/nfs1/home/USERNAME
  OSL Home Directory	/nfs/rontok/xraid/users/USERNAME
  Extreme Home Directory	/nfs/mneme/home/users/USERNAME

• /tmp and /var/tmp - All cluster nodes have the standard unix /tmp and /var/tmp directories. These are on the local disk on each node and are not accessible from any other node. Any files created in these directories should be removed at the end of your job and any files left behind will be automatically removed.

• /san01 and /san02 - There are two 4TB file systems mounted as /san01 and /san02 on tyr05, tyr06, tyr07, and tyr08 only. These filesystems are NOT BACKED UP.

• san36-1-0 - There is an 8Tb slice of the 36TB disk array accessible on the tyr nodes using the path /nfs/odinnfs1/san36-1-0. This filesystem is NOT BACKED UP.

Please see the Storage and Network Topology FAQ for more information about the storage and network topology of the clusters.

Please let us know if you have any questions about these storage option or if you need space allocated.

[Return to the FAQ index]