Is RAID a Backup? What It Protects You From—and What It Doesn’t

Maria Torres

Maria Torres

September 27, 2026

Is RAID a Backup? What It Protects You From—and What It Doesn't

The call came on a Monday morning from a wedding photographer I’d done some IT work for a couple of years earlier. Her assistant had been tidying up a laptop over the weekend and deleted what he thought was a duplicate folder. It wasn’t a duplicate. It was the synced copy of three weddings’ worth of edited images, and the sync client had faithfully deleted the originals from the studio’s NAS within a few seconds.

“But we have RAID,” she said. “Doesn’t that mean there’s a second copy?”

There was a second copy. There were, in fact, four drives’ worth of redundancy in that NAS. Every one of them had deleted the folder at the same instant, because that’s exactly what RAID is designed to do. We got most of the work back from a USB drive she’d copied things to about five weeks earlier. The five weeks of edits in between were gone.

I’ve heard “but we have RAID” more times than I can count, in homes and in businesses. It’s an understandable mistake, because RAID really does keep multiple copies of your data. It just keeps them for a very different reason than a backup does.

What RAID actually is

RAID stands for redundant array of independent disks. It combines several drives so that the system can survive one or more of them failing. There are different layouts, called levels, but the ones you’ll meet at home are:

  • RAID 1 (mirror): two drives hold identical data. One can die and the other carries on.
  • RAID 5 (or RAID-Z1): three or more drives, with parity spread across them. Any one drive can fail.
  • RAID 6 (or RAID-Z2): four or more drives with double parity. Any two can fail.
  • Synology SHR and Unraid’s parity are variations that handle mixed drive sizes, but the principle is the same.

The key word in all of these is live. Every write goes to all the drives at once. That’s the whole point. When one drive dies, the others already have everything, up to the millisecond, so your files stay available while you replace the dead one.

That same design is why RAID isn’t a backup. It copies everything instantly, including mistakes.

Worried woman at a monitor in a small photography studio office with a NAS on a filing cabinet

What RAID protects you from

Let’s give RAID its due. It solves real problems.

A single drive failing. This is the big one. Hard drives die. In any given year, a small percentage of drives in a large population fail, and the odds rise as they age. Without RAID, a dead drive means everything on it is gone until you restore from backup, which might take hours or days. With RAID, you get an email, order a replacement, swap it in, and nothing was ever offline.

Downtime. That’s really what RAID buys you: availability. Your media server keeps streaming, your photo library keeps loading, and your family doesn’t notice a drive died last Tuesday. For a business, that alone can justify it.

Some kinds of silent corruption, if your filesystem helps. Drives occasionally return bad data without reporting an error. Classic RAID can’t tell which copy is right. Filesystems that checksum every block, like ZFS and btrfs, can detect the bad copy and repair it from the good one during a scrub. That’s a meaningful extra layer, and it’s one of the main reasons people choose those filesystems for home servers. If you’re deciding between them, ZFS RAID-Z2 versus btrfs RAID on a single box covers how each handles that job.

That’s the list. It’s a good list. But notice that every item on it is about hardware failing. None of it is about you, your software, or the world around the box.

What RAID does not protect you from

Deleting something. You delete a file, RAID deletes it from every drive. Instantly. This is the photographer’s story, and it’s by far the most common way people lose data.

Overwriting something. Save over the wrong spreadsheet, let an app “clean up” a library, run a script with a typo in the path. RAID faithfully writes the damage to all drives.

Ransomware. Malware that encrypts your files does it through the same normal file operations as any other program. RAID sees a lot of writes and dutifully mirrors every encrypted byte. A NAS with a share mapped on an infected laptop is one of the most common ways home users and small offices lose everything.

Software and filesystem bugs. If an app corrupts its own database, or the filesystem itself gets damaged, RAID doesn’t help, because the corruption happens above the level RAID works at. Checksumming filesystems catch bad data from the drive. They can’t catch bad data that an application wrote on purpose.

The box dying. A power supply failure, a lightning surge, a motherboard that fries itself: all of these can take multiple drives with them, or leave you with an array that only the original hardware knows how to read. Software RAID like Linux mdadm, ZFS, and Synology’s SHR can usually be recovered in another machine. Some hardware RAID controllers and proprietary formats are much harder.

Fire, flood, and theft. All the drives are in the same box, in the same room. Whatever happens to the room happens to every copy.

You making a mistake during a repair. I’ve seen someone pull the healthy drive out of a degraded mirror instead of the dead one. RAID is not forgiving about that.

The rebuild problem nobody warns you about

There’s one more failure mode that deserves its own section, because it catches people who think they’re safe.

When a drive in a RAID 5 array dies, the array keeps running with no redundancy left. You replace the drive, and the system starts a rebuild: it reads every block on every remaining drive to reconstruct the missing one. On large modern drives, that can take a day or more, and it’s the most stressful thing those drives will ever do.

If a second drive fails during the rebuild, or throws enough read errors, the array can be lost. That’s more likely than it sounds. Drives in a home NAS were often bought together, from the same batch, and have run the same hours in the same heat. When one wears out, the others aren’t far behind.

Two years after the photographer’s call, another client lost a four-drive RAID 5 exactly this way. One drive failed, the replacement went in, and about sixteen hours into the rebuild a second drive started returning errors. That array had no backup at all. It went to a recovery lab and cost more than the NAS and all its drives combined.

This is why most people who care about the question now recommend double parity, like RAID 6 or RAID-Z2, or mirrors, for anything over a few terabytes. But even RAID 6 is still RAID. It narrows one risk. It does nothing about the other list.

Hand pulling a drive from a two-bay NAS with a replacement drive in an anti-static bag beside it

What about snapshots?

Snapshots are the thing people bring up next, and they’re genuinely useful. ZFS, btrfs, and most modern NAS systems can take a snapshot of a volume in a second, storing only the blocks that change afterward. If someone deletes a folder, you can roll it back from an hour ago. If ransomware encrypts a share, you can often restore the whole share to before it started.

If the photographer’s NAS had taken hourly snapshots, her Monday would have been a five-minute fix. Snapshots cover the “oops” category that RAID ignores, and every NAS owner should turn them on.

But snapshots live on the same drives, in the same box, in the same room. When the box dies, the building floods, or the array fails during a rebuild, the snapshots go with it. Some ransomware strains also go looking for snapshots to delete if they get administrator access to the NAS. Snapshots are a very good undo button. They’re still not a backup.

What a backup actually is

A backup is a separate copy of your data that doesn’t share the same failure. If the thing that destroys your original can also destroy the copy, it isn’t a backup.

The long-standing rule of thumb is 3-2-1: three copies of your data, on two different kinds of storage or devices, with one of them offsite. For a home server, that usually looks like:

  • The original on your NAS or server, on RAID if you want uptime.
  • A second copy on a different device: an external drive, a second small NAS, or an old PC.
  • A third copy offsite: cloud storage, a drive at a relative’s house, or a NAS at a friend’s place.

Two properties matter as much as the count. Backups should be versioned, so you can go back to last week or last month, not just last night, because it can take weeks to notice a deletion. And at least one copy should be hard for your main system to change. That can be a drive you unplug, cloud storage with object lock, or a backup server that pulls data rather than letting the NAS push and delete. If ransomware on your laptop can reach the backup, it can encrypt the backup.

Finally, a backup you’ve never restored from is a hope, not a backup. Pick a random folder once a quarter and actually restore it somewhere. You’ll either sleep better or find a problem while it’s still cheap.

So, do you still need RAID?

Maybe not. This surprises people, but if you have good backups and can tolerate a day of downtime while you restore, a home server without RAID is a perfectly reasonable choice. You get more usable space from the same drives and one less thing to manage.

RAID earns its place when downtime matters more than disk space: a family photo library everyone uses every day, a media server with several users, a small business NAS. Then it’s worth having, as long as everyone understands what it’s for.

The way I explain it to clients now is three layers, each covering a different kind of bad day:

  • RAID keeps you running when a drive dies.
  • Snapshots let you undo mistakes quickly.
  • Backups save you when the box, the building, or the malware wins.

The photographer has all three now. Her NAS takes hourly snapshots, backs up nightly to a second unit at her home, and sends a weekly encrypted copy to cloud storage with object lock. Last spring the same assistant deleted a different folder. She restored it from a snapshot before he’d finished apologising.

More articles for you