What Is RAID? RAID Levels Explained (0, 1, 5, 6, 10)
If you’ve bought a NAS, set up a server, or read a spec sheet for a storage box, you’ve run into the term. So what is RAID? RAID stands for Redundant Array of Independent Disks (originally “Inexpensive”). It combines several physical drives into one logical volume to gain speed, survive a drive failure, or both. Which of those you get depends on the RAID level, and the trade-offs between levels matter a lot more than most setup wizards let on. Here’s how it works, with the main RAID levels explained side by side.
What is RAID, in plain terms?
Every RAID level is built from three ideas:
- Striping: data is split into chunks and spread across several drives, so multiple drives work on a request at once. Faster, but on its own it offers no protection.
- Mirroring: the same data is written to two (or more) drives. If one dies, the copy survives.
- Parity: extra calculated data that lets the array rebuild the contents of a missing drive from what’s on the others. It protects you while using less space than a full mirror.
The operating system sees one big drive. Underneath, a RAID controller or software layer decides where each chunk goes. That’s why individual drives from an array usually can’t be read on their own: each one holds only pieces of the whole.
RAID levels explained
RAID 0 (striping)
Data is striped across two or more drives with no redundancy. You get the full capacity of every drive and the best raw performance, but if any single drive fails, the whole array is lost. RAID 0 suits scratch space and temporary video work, never the only copy of anything. More on recovery in our RAID 0 recovery page.
RAID 1 (mirroring)
Two drives hold identical copies. Either one can fail and the data survives. You only get the capacity of one drive, and write speed is roughly that of a single drive. RAID 1 is simple and common in small servers and two-bay NAS units.
RAID 5 (striping with parity)
Data and parity are striped across three or more drives, with the parity rotated among them. The array survives one drive failure, and you lose one drive’s worth of capacity to parity. It’s a popular balance of space and protection, but rebuilds on large drives are long and stressful. See RAID 5 vs RAID 6 for why that matters.
RAID 6 (dual parity)
Like RAID 5, but with two independent sets of parity. It needs at least four drives, gives up two drives’ worth of capacity, and survives any two drives failing. Writes are slower than RAID 5 because there’s more parity to calculate.
RAID 10 (mirrored pairs, striped)
Drives are paired into mirrors, then data is striped across the pairs. It needs at least four drives and gives you half the total capacity. Performance is excellent, especially for random writes, and rebuilds only copy from a mirror partner. It can survive multiple failures, but only if no mirror pair loses both drives. Our RAID 10 vs RAID 5 guide compares the two in depth.
JBOD, RAID 50 and RAID 60
- JBOD (“just a bunch of disks”) usually means drives presented individually or spanned end to end into one volume. There’s no redundancy; a failed drive takes its portion of the data with it, and depending on the file system it can make the rest hard to access.
- RAID 50 and RAID 60 stripe data across several RAID 5 or RAID 6 groups. They’re used in larger servers to get better performance and shorter rebuilds than one giant RAID 5 or 6. RAID 50 needs at least six drives; RAID 60 at least eight.
RAID levels compared
| Level | Min. drives | Survives | Usable capacity | Performance |
|---|---|---|---|---|
| RAID 0 | 2 | No drive failures | 100% of total | Fastest reads and writes |
| RAID 1 | 2 | All but one drive in the mirror | One drive’s worth | Good reads, single-drive writes |
| RAID 5 | 3 | 1 drive | Total minus 1 drive | Good reads, slower small writes |
| RAID 6 | 4 | Any 2 drives | Total minus 2 drives | Good reads, slowest small writes |
| RAID 10 | 4 | 1 drive per mirror pair | 50% of total | Fast reads and writes |
| JBOD / span | 2 | None | 100% of total | Same as a single drive |
Hardware vs software RAID
Hardware RAID uses a dedicated controller card (or a controller built into a server) with its own processor and often a battery- or flash-backed cache. The operating system just sees the finished volume. It’s fast and OS-independent, but the array’s configuration lives in the controller’s own format, so if the controller dies, you typically need a compatible controller to bring the array back.
Software RAID is handled by the operating system or storage platform: Linux mdadm, ZFS, Windows Storage Spaces, macOS Disk Utility RAID sets, and most NAS operating systems. It’s flexible and doesn’t depend on one specific card. Motherboard “RAID” (such as Intel RST) sits in between: configured in firmware, but relying on a driver in the operating system.
Vendor schemes: SHR, BeyondRAID and unRAID
Some NAS makers layer their own schemes on top of standard RAID, mainly to allow mixed drive sizes:
- Synology Hybrid RAID (SHR) is built on standard Linux RAID and volume management. SHR-1 tolerates one drive failure and SHR-2 tolerates two, roughly like RAID 5 and RAID 6, while making better use of mismatched drives.
- Drobo BeyondRAID is a proprietary scheme that manages redundancy across drives of different sizes. Drobo units are no longer sold new, but plenty are still running.
- unRAID works differently: each data drive generally keeps its own complete file system, with one or two dedicated parity drives protecting them. Lose more drives than you have parity, and you lose what was on the failed drives rather than necessarily the whole array.
Exact behavior varies by version and configuration, so check your own system’s documentation before assuming how many failures it can take.
RAID is not a backup
This is the single most important thing to understand about RAID. Redundancy protects you from a drive failing. It does nothing about:
- Deleting or overwriting a file (the deletion is mirrored or parity-protected instantly)
- Ransomware, which encrypts the array just as happily as a single drive (see ransomware file recovery)
- File system corruption, controller faults, firmware bugs, or a failed rebuild
- Fire, theft, flooding, or a power surge that damages every drive at once
- Several drives from the same batch wearing out around the same time
A real backup is a separate copy on separate hardware, ideally with one copy offsite. The 3-2-1 backup rule is the simplest way to get there. Treat RAID as uptime insurance, not data insurance.
What happens when a RAID fails
Most RAID data loss follows a familiar pattern. One drive fails and the array keeps running in a degraded state. Nobody notices, or someone replaces the drive and starts a rebuild. Then a second drive fails or hits unreadable sectors mid-rebuild, and the array goes offline. Other common causes are a dead controller, a NAS that won’t boot after an update, or someone reinitializing the array while trying to fix it.
If that’s where you are, stop forcing the array online, label every drive’s slot, and don’t reinitialize or run repair tools. Our RAID failure guide covers the details, and for a degraded array that still works, our guide to RAID 5 degraded and failed rebuilds walks through the safe order of operations.
How RAID recovery works
Because data is spread across the members, recovery means putting the array back together. Engineers image every drive individually (repairing any with mechanical faults first), then work out the array’s geometry: drive order, stripe size, parity rotation, and offsets. The array is then rebuilt virtually from the images, without writing to your original drives, and the files are extracted. Even when the controller or NAS is dead, the data on the drives is often intact.
The bottom line
RAID combines drives to gain speed, redundancy, or both. RAID 0 is fast with no protection; RAID 1 mirrors; RAID 5 and 6 use parity to survive one or two failures; RAID 10 mirrors and stripes for speed and quick rebuilds. Pick the level that fits your workload, but never confuse redundancy with backup.
If an array has already failed, our RAID data recovery service reconstructs every common RAID level, NAS scheme, and controller type by mail-in. Each case gets a custom quote after a free evaluation, and you pay nothing unless we recover your data.