To my surprise the Wikipedia on this subject is fairly poor (at the time of writing).
Journaling file system - Wikipedia, the free encyclopedia
As is the entry on Veritas (a popular commercial implementation).
Veritas File System - Wikipedia, the free encyclopedia
Hard discs fail. Mains power fails. Any commercial system designer must take this into account. You must always be in a position to recover from a hard disc failure or a mains power failure. Offsite backups cover you for major disaster (fire, flood etc.) but day-to-day disc failure should not interrupt processing and even a mains power failure should not corrupt your discs.
To cover this eventuality you need a Journalling Filesystem and sufficient RAID or Mirror (or preferably both) disc arrays to cover you against failure.
A half-decent Journalling Filesystem (like Veritas) tries to always maintains the visible (to software) discs in a valid state. This does not mean that you might not have to restore a backup, but that you will not have to rebuild the computer after a failure.
A single disc with Journalling Filesystem comes into the "Chocolate Teapot" category. You always need your data triplicated (or better) such that the Journalling Filesystem software can determine "WHICH IS THE BAD DISC?" and request its replacement. In the ideal world you should be able to keep running 7/24 given an infallable UPS and a 100% sound policy on replacing failed hard discs.
A recent post on this forum showed that all these precautions are worthless unless you have an alert mechanism for failed hard discs and a local process for reacting to alerts such that you immediately replace failed hard discs.
Main advantage of Journalling Filesystems:
It is possible to configure the computer to survive the common occurrence of single hard disc failure. Most importantly it is possible to configure a system to determine which is the "bad disc".
Main disadvantage of Journalling Filesystems:
Budgetary or irrational design restraints can give a wrong expectation of resilience and cause designers to be complacent about Disaster Recovery backups.
Footnote: If I have critical data on a system, the mininum disc configuration is RAID-5 Triple-mirrored. I need to be able to survive the failure of a RAID controller - not just a single disc. There are modern SANs which provide this level of hardware resilience.