Backup Policy

Backup Policy

TIG-managed AFS and NFS storage space¹ is backed up on the following schedule:

• Full backups (level 0)
  • Taken upon creation and population
  • Mirrored tape writes
  • One copy stored offsite for one year
• Monthly backups (level 1)
  • Taken monthly
  • Mirrored tape writes
  • One copy stored offsite for one year
• Weekly backups (level 2)
  • Taken weekly
  • Archived for one month
• Daily backups (incremental)
  • Taken daily
  • Archived for two weeks

Additionally, AFS storage is snapshotted nightly. A copy-on-write snapshot of each volume is available, however only the AFS home directory has the snapshot mounted by default (as ~/.snapshot). If you need the snapshot feature and your AFS project directory or AFS group directory does not have a .snapshot subdirectory please email help@csail.mit.edu and we can set this up for you.

As can be seen from this schedule, the monthly backup is the only level of granularity guaranteed over time. For requests of data less than two weeks old we can usually provide single-day resolution. Between two and four weeks, we can only resolve to the week level. Past that it’s monthly. We archive backup tapes as long as possible, but we cannot guarantee the physical integrity of the tapes in excess of one year.

The somewhat odd four level backup system with mirrored level one backups allows us to conserve tape since we only write out full data once and the monthlies tend to be about 20% this size. The mirrored monthlies allow us to lose any one tape without losing data. A lost full can be reconstructed from previous monthly and full tapes, and there are two copies of each monthly.