RAID levels explained: RAID 1 vs. RAID 5 vs. RAID 10
Respuesta rápida
RAID protects against drive failure by spreading or duplicating data across multiple disks, and the level you pick trades off capacity, performance and rebuild risk. RAID 1 mirrors two drives and is the simplest, safest choice for boot drives or small setups. RAID 5 uses parity for better capacity efficiency but carries growing rebuild risk on large modern drives. RAID 10 stripes across mirrored pairs and gives the best combination of performance and redundancy, at the cost of needing more drives.
What RAID actually does
RAID (Redundant Array of Independent Disks) combines multiple physical drives into one logical unit, so the array can survive a drive failure without losing data, and in most configurations without taking the server offline. Different RAID levels achieve that in different ways: by mirroring the same data across drives, by writing parity information alongside the data, or by combining mirroring and striping together. That choice of technique is what drives the differences in capacity overhead, performance and rebuild behaviour below.
RAID protects against a drive failing. It does not protect against accidental deletion, ransomware, a bad software update, or the array controller itself failing, all of the data stays lost if the array is destroyed or corrupted as a whole. RAID and backups solve different problems, and one is not a substitute for the other.
RAID 1: mirroring
RAID 1 writes the same data to two drives simultaneously. If one drive fails, the other keeps serving data without interruption, and the failed drive can be replaced and rebuilt from the surviving copy. It's the simplest RAID level to understand and to recover from, which is exactly why it's a common default for boot drives and small two-drive setups.
The trade-off is capacity: with two drives in RAID 1, you get the usable capacity of one drive, a 50% overhead. Write performance is roughly that of a single drive, since every write has to complete on both mirrors, but read performance can be better than a single drive since reads can be served from either side of the mirror.
RAID 5: parity
RAID 5 spreads data across three or more drives along with a parity block that can reconstruct any one missing drive's data. That gives noticeably better capacity efficiency than mirroring: with three drives, you get the usable capacity of two, and the ratio improves as more drives join the array.
The catch is what happens when a drive actually fails. Rebuilding a RAID 5 array means reading every remaining drive in full to reconstruct the failed one, and on today's large-capacity drives that rebuild can take a long time, during which the array has no redundancy left and a second drive failure or an unrecoverable read error can mean data loss. That risk has grown as drive capacities have grown, which is why RAID 5 is increasingly being phased out in favour of RAID 10 or RAID 6 for anything where the rebuild window matters.
RAID 10: striped mirrors
RAID 10 (sometimes written RAID 1+0) combines both techniques: drives are first mirrored in pairs, then those mirrored pairs are striped together. The result reads and writes across multiple mirrors at once, giving performance closer to plain striping, while still tolerating a drive failure in each mirrored pair independently.
This combination is why RAID 10 is the most common default for database and VM workloads that need both speed and redundancy: rebuilds only need to copy from the surviving half of one mirror, not read the whole array, so they're faster and safer than a RAID 5 rebuild. The trade-off is drive count and capacity: RAID 10 needs at least four drives, and usable capacity is half the total raw capacity, the same 50% overhead as RAID 1, just spread across more drives for more performance.
Comparing the three
| RAID level | Minimum drives | Capacity overhead | Fault tolerance | Read/write performance | Best for |
|---|---|---|---|---|---|
| RAID 1 | 2 | 50% | 1 drive per mirrored pair | Good reads, write speed of one drive | Boot drives, small two-drive setups |
| RAID 5 | 3 | 1 drive's worth (improves with more drives) | 1 drive, array is exposed during rebuild | Good reads, parity calculation slows writes | Capacity-focused setups on smaller drives, being phased out for large drives |
| RAID 10 | 4 | 50% | 1 drive per mirrored pair, can survive more depending on which drives fail | Best of the three for both reads and writes | Databases, VM hosts, workloads needing performance and redundancy together |
Where this fits when ordering
The RAID level applied to a dedicated server is shown as part of the storage configuration at order time, alongside the drive setup itself, see Configuring a dedicated server when you order it. If you're running Windows and configuring RAID at the OS level rather than through hardware, see Windows Software RAID: 0, 1, and 5 for how that works in practice.