Short answer: RAID (Redundant Array of Independent Disks) is a storage method that makes several physical disks work as one logical disk. Depending on the level, it gives you speed (RAID 0), protection against disk failure (RAID 1) or a balance of both (RAID 5, 6, 10). Servers most often use RAID 1 and RAID 10. Important: RAID is not a backup; it only prevents downtime when a disk fails.
How Does RAID Work?
The RAID concept was defined in a 1988 paper by researchers at the University of California, Berkeley (back then the “I” stood for “Inexpensive”). Every RAID level is a different combination of three basic techniques:
- Striping: data is split into blocks and written across the disks in turn. Because the disks work in parallel, speed goes up, but striping alone offers no protection.
- Mirroring: the same data is written identically to two or more disks. If one disk fails, its copy takes over.
- Parity: checksum information calculated from the data blocks. When a disk is lost, the missing data is recalculated from the remaining data and the parity.
When a failed disk is replaced, RAID writes the missing data onto the new disk. This is called a rebuild; on large disks it can take hours, and the system slows down while it runs.
RAID Levels
RAID 0: Striping (Speed Only)
Data is split across all disks; read and write speed grows with the number of disks and the full capacity is usable. But there is no protection at all: if a single disk fails, all data in the array is lost. It suits jobs where the data already exists elsewhere, such as scratch space for video editing or a cache.
RAID 1: Mirroring (Safety)
All data is written identically to at least two disks. In a two-disk RAID 1, the system keeps running even if one disk fails. Reads can be faster because they can come from both disks; write speed is close to a single disk. The cost is capacity: two 1 TB disks give you 1 TB of usable space. It’s the most common choice for operating system disks and small servers.
RAID 5: Distributed Parity
Requires at least three disks. Data and parity are spread across all disks, and one disk’s worth of capacity goes to parity. For example, four 2 TB disks give 6 TB of usable space. It survives a single disk failure. Reads are fast, but because parity is updated on every write, write performance drops. On large-capacity disks a rebuild takes a long time, and a second disk problem during it can lose the whole array; that’s why RAID 6 or RAID 10 is usually preferred for large arrays.
RAID 6: Double Parity
Requires at least four disks. It’s similar to RAID 5 but keeps two separate sets of parity, so it survives two disk failures at the same time. Usable capacity equals the total number of disks minus two. Write performance is even lower than RAID 5. It’s used for large, read-heavy arrays such as archives and backup storage.
RAID 10 (1+0): A Stripe of Mirrors
Requires at least four disks. Disks are first mirrored in pairs (RAID 1), and the pairs are then striped (RAID 0). With no parity calculation, writes are fast and rebuilds are quick. One disk from each mirrored pair can fail; but if both disks in the same pair fail, data is lost. Half the capacity is usable. It’s the level most often recommended for database and virtualisation servers.
RAID Levels Comparison Table
| Feature | RAID 0 | RAID 1 | RAID 5 | RAID 6 | RAID 10 |
|---|---|---|---|---|---|
| Minimum disks | 2 | 2 | 3 | 4 | 4 |
| Usable capacity | All (N) | One disk | N − 1 disks | N − 2 disks | Half |
| Failures survived | None | 1 disk (in a 2-disk array) | 1 disk | 2 disks | 1 disk per pair |
| Read speed | Very high | High | High | High | Very high |
| Write speed | Very high | Like a single disk | Medium | Low–medium | High |
| Typical use | Scratch space, cache | OS disk, small server | File server | Archive, large storage | Database, virtualisation |
Software RAID or Hardware RAID?
| Software RAID | Hardware RAID | |
|---|---|---|
| How it works | Managed by the operating system (mdadm on Linux, Storage Spaces on Windows) | Managed by a separate RAID controller card |
| Cost | No extra hardware needed | The controller card is an extra cost |
| Performance | Enough for most workloads on modern CPUs | Cards with cache can give better write performance |
| Portability | Disks can be moved to another Linux system and read | If the card fails you usually need the same make/model of card |
On Linux you can check software RAID status with these commands:
# Summary of all RAID arrays and rebuild progress
cat /proc/mdstat
# Details of a specific array
sudo mdadm --detail /dev/md0RAID Is Not a Backup
RAID only protects against disk failure. An accidentally deleted file, data encrypted by ransomware or a corrupted database is reflected on every disk at once. That’s why you always need a separate backup plan alongside RAID. See our server backup strategies article for details.
Which RAID Level Should I Choose?
- 2 disks, safety first: RAID 1
- Database or virtualisation, 4+ disks: RAID 10
- Many large disks, archive or file store: RAID 6
- Speed only, data backed up elsewhere: RAID 0
If you’re looking for a server where you can choose your own disk layout, take a look at our dedicated server options.
Frequently Asked Questions
Is RAID 1 or RAID 5 better?
With two disks, RAID 1 is the only option. With three or more disks RAID 5 gives more capacity, but RAID 1 and RAID 10 are considered safer in terms of write performance and rebuild risk.
Can you use disks of different sizes in RAID?
You can, but in classic RAID every disk counts as the size of the smallest disk in the array; the extra capacity is wasted. That’s why disks of the same size and model are recommended.
Can you use RAID with SSD or NVMe drives?
Yes. The same RAID levels are used with SSD and NVMe drives; rebuilds on SSDs are much shorter than on HDDs.
Conclusion
RAID combines several disks to deliver speed, fault tolerance or both. RAID 0 is speed only, RAID 1 is simple and safe mirroring, RAID 5 and 6 are capacity-focused parity, and RAID 10 balances speed and safety. Whichever you choose, put RAID alongside regular backups, not in place of them.
Categories
-
CentOS 7 guides
-
cPanel 20 guides
-
CyberPanel 18 guides
-
Operating Systems 81 guides
-
Juniper Networks 18 guides
-
Control Panels 50 guides
-
Linux 17 guides
-
MySQL 29 guides
-
Network 95 guides
-
Plesk Panel 12 guides
-
Virtual Servers 273 guides
-
SEO 3 guides
-
Cyber Security 8 guides
-
Technology 162 guides
-
VMware 36 guides
-
WordPress 20 guides
-
Artificial Intelligence 22 guides