RAID can save you from a failed disk. Replication can save you from a failed machine. Neither necessarily saves you from deleting the wrong directory.
That distinction matters, because availability and recoverability are different problems and the word "redundant" is routinely used as though it solved both.
RAID keeps storage available
Depending on the layout, data is spread or duplicated across several disks. When one fails, the system can usually keep running while the device is replaced. That is genuinely useful, because disks fail.
But if ransomware encrypts a file, RAID faithfully stores the encrypted file. If an administrator drops a database, RAID faithfully records the deletion.
It protects against some hardware failures. It does not protect the logical contents from anything at all.
Replication keeps another copy current
Replication sends changes to another system, which can give excellent availability: if a host or a site fails, another already holds an up-to-date copy.
The whole problem is contained in the phrase "up-to-date". If the current state is wrong, replication makes the second copy wrong very efficiently.
Delete a table and the deletion replicates. Corrupt a record and the corruption replicates. Compromise a credential and the attacker may well reach both sides.
Snapshots add history
Snapshots improve the picture, because they preserve previous states. They are excellent for quick recovery from operational mistakes and give useful short-term history.
But a snapshot usually lives inside the storage system it is protecting. It can disappear with the array, the account or the site, and some are reachable by exactly the administrative credentials an attacker would target first.
A recovery tool, then, but not on its own a backup strategy.
Backups deliberately preserve recoverable history
A backup system is designed around recovery from data loss rather than around continuity of service. It keeps versions according to a retention policy, and it stores copies across different failure domains — with at least one that normal production credentials cannot rewrite or delete.
That is what lets you say the sentence that matters: the current state is broken, give me the state from Tuesday morning.
Use all of them for what they are good at
These are not competitors, and a well-designed system may use all four.
- RAID handles disk failure.
- Replication handles system or site failure.
- Snapshots give fast short-term rollback.
- Backups give independent historical recovery.
The mistake is buying one of them and assuming the data is now safe.