RAID combines several drives into one array for capacity, speed or redundancy — but every level trades those three against each other differently. This guide explains how each level works, what it costs in capacity, and how many drives it can lose, with a live RAID calculator to size your own.
What RAID is (and is not)
RAID — a redundant array of independent disks — presents several physical drives to the operating system as one logical volume. Depending on the level, it spreads data for speed (striping), keeps copies for safety (mirroring), or stores parity that can rebuild a failed drive. The goal is to survive a drive failure and/or go faster than a single disk.
What RAID is not is a backup. It protects against drive failure, not against deletion, corruption, ransomware or site loss — those need a separate, ideally air-gapped copy. We cover that distinction in RAID is not a backup.
The three building blocks: striping, mirroring, parity
Striping (RAID 0) splits data across all drives for maximum speed and 100% capacity, but any one drive failure loses everything. Mirroring (RAID 1) keeps a full copy on each drive — safe and fast to rebuild, but only 50% efficient. Parity (RAID 5/6) stores a mathematical checksum so a failed drive can be reconstructed, keeping most of your capacity at the cost of a write penalty.
Nested levels combine these: RAID 10 stripes across mirror pairs (speed + safety), while RAID 50/60 stripe across parity groups (capacity + faster rebuilds). ZFS RAIDZ is parity done in software with checksums and self-healing built in.
Capacity, fault tolerance and write penalty by level
Usable capacity follows simple formulas: RAID 0 is n × drive size; RAID 5 is (n−1) × size; RAID 6 and RAID-DP are (n−2) × size; RAID 10 is n/2 × size; RAID 50/60 multiply per parity group; and RAIDZ removes parity per vdev. The calculator applies the right one and shows the parity overhead.
Fault tolerance is what you can lose: 0 for RAID 0, one drive for RAID 5/RAIDZ1, two for RAID 6/RAIDZ2/RAID-DP, three for RAIDZ3/RAID-TEC. The write penalty — extra back-end I/O per write — is ×1 for RAID 0, ×2 for RAID 1/10, ×4 for RAID 5, and ×6 for RAID 6, which is why write-heavy databases prefer RAID 10.
Picking a level
For boot drives, use RAID 1. For write-heavy databases and VMs, use RAID 10. For large capacity HDD pools, use RAID 6 (or RAID 60 at scale) — single-parity RAID 5 is risky on big drives because a rebuild has no safety net, as explained in is RAID 5 dead?. For TrueNAS/ZFS, RAIDZ2 is the wide-vdev default.
There is no single best level — only the best fit for your capacity, performance and resilience needs. Use the calculator to compare two levels on your exact drives, then talk to us about the array to run it.