RAID is not a backup

RAID combines several disks into one volume, and many people assume that this makes their data safe. It does not, and the difference between redundancy and backup is worth understanding before a disk fails.

RAID 1 mirrors the same data on two disks, so one can die without data loss. RAID 5 spreads data and parity over at least three disks and survives one failure, while RAID 6 survives two. RAID 10 combines mirrors and striping for speed and resilience. RAID 0, by contrast, splits data across disks with no redundancy at all, so losing one disk loses everything.

All of these protect against one thing: a disk dying. They do nothing against accidental deletion, because a deleted file is deleted on every disk at once. They do not help when a file is corrupted or encrypted by ransomware, since the damage is copied instantly. A failing controller, a power surge, a fire or a theft can take out the whole array together.

Rebuilding is also a risky time. When a disk is replaced, the array reads all the remaining disks heavily for hours or days, and if another disk fails during that period, a RAID 5 array can be lost. Disks bought together often have the same age and wear, which raises the chance of a second failure.

The sensible setup is both: RAID for availability, so the system keeps running when a disk dies, and separate backups for recovery, stored on a different device and ideally in a different place. If you can afford only one, choose the backup.