HPC Scratch Space

Scratch Space Described

Scratch space is temporary storage for files.  Most of the time you will want to use your own permanent storage but there are situations where scratch might be more appropriate.  For example,

  • You are doing a computation that requires large amounts of temporary space but only while the computation is running.
  • You have a data set larger than what you can fit into your own permanent storage but you only need it for less than 45 days.
  • You are doing a computation that needs higher performance storage than what your own permanent storage can provide (mostly applicable to local SSD scratch drives).

Scratch policies

  • Don’t write directly to the top level of a scratch volume. Instead create a sub-directory that matches your username. For example “mkdir /scratch/mynetid” if you login with your netid and want to use the /scratch volume on one of our systems.
  • Files and directories older than 45 days in scratch are automatically purged because this is a shared and limited resource.
  • One week before any of your files are deleted you will receive a warning email. When possible, we include a list of those files. This is your opportunity to request an extension for files whose usefulness has gone on longer than expected.
  • We cannot make any guarantees about the safety of data in scratch volumes. It’s intended to be temporary and safety is traded off for performance. There are no backups and many of these volumes are arrays of disks where failure of one disk means the loss of all data in the volume.

Characteristics of specific scratch volumes


  • network storage, available to entire Discovery cluster plus Polaris and Andes
  • exact same performance as anything else mounted from DartFS (e.g. home and lab directories)
  • not subject to data loss from drive failure


  • local to that machine
  • newer nodes have SSD and performance will be noticeably better than /dartfs-hpc/scratch for random I/O
  • older nodes have spinning disks and performance is probably around the same as /dartfs-hpc/scratch 
  • subject to data loss with a drive failure


  • local to Polaris
  • an extremely fast array of drives
  • excellent for streaming reads/writes
  • subject to data loss with individual drive failure


  • local to Andes
  • an extremely fast array of drives
  • excellent for streaming reads/writes
  • subject to data loss individual drive failure


  • DEPRECATED and will be merging with /dartfs-hpc/scratch in summer of 2023.
Print Article


Article ID: 140938
Wed 1/26/22 12:14 PM
Fri 8/4/23 12:22 PM