Backing-up filesystems

Although there are several ways to reduce or eliminate corruption of the filesystem, practically one must back-up valuable data as frequently as possible. There are three main back-up approaches:

1. Full dump -- Make an entire copy of the filesystem on another disk, or on tape. In the event of the filesystem being damaged (or even someone accidentally deleting a file), copy across all or some of the backed-up filesystem.
2. Incremental -- Only backup the files which have changed. Some systems use an 'archived' bit associated with each file, which is cleared when the file is created or modified, and set when the file is backed-up, that way the backup software can tell if the file is due to be backed-up, or whether it has not changed, and need not be backed-up.

   Other systems (e.g. UNIX does not assign an 'archived' bit to files) must store a list of files and when they were last backed up. It can then check the 'last backed-up' time against the 'last modified' time, to determine whether they have changed and must be written to the backup medium.

3. Journalling -- More incremental still is the idea of logging all file operations, as mentioned above. Trouble with this strategy generally is that many small alterations may be made to each file, and in the long-run incremental scheme is more efficient with backup space.

Many computer system installations perform a regular incremental backup (say once a day), then take a full dump at the end of the week, or end of the month. Could imagine a three-level strategy, using all three techniques: even with a sophisticated journalling system, it is important to take a full dump regularly.