Skip to content

Guide / Storage

Choosing a RAID level for your server

RAID 0, 1, 5, 6 and 10 explained: usable capacity, fault tolerance and performance trade-offs, and how to pick the right layout.

RAID (Redundant Array of Independent Disks) combines several drives into one logical volume. Depending on the level, it adds speed, redundancy or both — always at a cost in usable capacity. Here's how the common levels compare.

RAID 0 — striping

Data is split across all drives. You get the full capacity and the fastest throughput, but no redundancy: if any drive fails, the whole array is lost. Use it only for data you can easily recreate, such as scratch space.

RAID 1 — mirroring

Two drives hold identical copies. Usable capacity is one drive; the array survives the failure of either. Simple and reliable — a common choice for operating-system volumes.

RAID 5 — single parity

Data and parity are distributed across at least three drives, using one drive's worth of capacity for parity. It survives one drive failure. With today's large drives, rebuilds take a long time and put stress on the remaining drives, so a second failure during a rebuild is a real risk.

RAID 6 — double parity

Like RAID 5 with two drives' worth of parity (minimum four drives). It survives any two drive failures, which makes it a common choice for large-capacity arrays. Writes are slower than RAID 10.

RAID 10 — mirrored stripes

Drives are mirrored in pairs, and the pairs are striped together (minimum four drives). Half the raw capacity is usable. Write performance is strong and rebuilds are fast; it survives one failure per mirrored pair. Popular for databases and other write-heavy workloads.

Comparison

LevelMin. drivesUsable capacity (n drives)SurvivesWrites
RAID 02nNo failuresFastest
RAID 121 drive1 failureGood
RAID 53n − 11 failureModerate
RAID 64n − 22 failuresSlower
RAID 104n ÷ 21 per mirrored pairFast

Try these trade-offs interactively on the Storage Servers page.

Hardware or software RAID?

Hardware RAID uses a dedicated controller; software RAID (such as Linux mdadm) or RAID-aware file systems (such as ZFS) are managed by the operating system. Software approaches are flexible and transparent; hardware controllers can offload work from the CPU. Which options apply depends on the server configuration.

RAID is not a backup

RAID protects availability when a drive fails. It does not protect against deleted files, corruption, ransomware or the loss of the whole server — those changes are written to every drive instantly. Always keep separate, versioned backups: see Planning backups with the 3-2-1 rule.

Not sure which level suits your data? Ask sales@theclouders.com.

Need help with this on your server?

Email support