From Montana Tech High Performance Computing
The storage system on the cluster consists of 4 separate components:
Management node local storage - 450 GB RAID 1 (mirrored drives) The operating system and old users' home accounts are located on these drives. Backups are made to the NFS storage system - /data and /data1
Main storage system - 25 TB and 66TB RAID 5 and NFS storage accessed through the InfiniBand Network. The system includes a hot-swappable spare drive if a drive fails, the system will automatically replace the bad drive with it. This nfs0 storage node is the /data partition and nfs1 storage node is the /data1 partition where larger data set should be stored. The system administrator will setup a user or project account upon request. Newly created users' home directory will be on either of the NFS storage node.
Compute node memory - when booted, a lightweight version of the OS is loaded into main memory (RAM). The /tmp partition is also in memory and all files stored here will be lost when a node reboots.
Compute node scratch storage - each node has a single 450 GB drive for temporary storage of data that is not backed up. This is persistent storage if the node is rebooted. Files will be removed by the administrator if the drive fills.