Storage
RAID levels explained: capacity, fault tolerance and write penalty
What RAID 0, 1, 5, 6, 10, 50 and 60 give you in usable space, drive failures survived and random-write performance.
GIRIVON Engineering · Updated 19 Nov 2026 · 2 min read
RAID combines drives for capacity, performance and resilience. The trade-offs come down to three numbers: usable capacity, failures tolerated and the write penalty (disk operations per logical random write).
| Level | Usable drives (n total) | Failures tolerated | Write penalty |
|---|---|---|---|
| RAID 0 | n | 0 | 1 |
| RAID 1 | 1 (n-way mirror) | n − 1 | 2 |
| RAID 5 | n − 1 | 1 | 4 |
| RAID 6 | n − 2 | 2 | 6 |
| RAID 10 | n ÷ 2 | 1 guaranteed | 2 |
| RAID 50 | groups × (per group − 1) | 1 per group | 4 |
| RAID 60 | groups × (per group − 2) | 2 per group | 6 |
Worked example: twelve 20 TB drives
- RAID 6: 10 data drives → 200 TB usable (about 182 TiB), survives any two failures.
- RAID 10: 6 mirrored pairs → 120 TB usable, the best random-write performance.
- RAID 5: 220 TB usable, but only one failure tolerated.
Rebuild time is the hidden risk
Rebuilding a failed drive reads every other drive in the set. At 100 MB/s, a 20 TB drive takes about 55 hours to rebuild (20 × 10¹² bytes ÷ 10⁸ bytes/s). During that window a single-parity array has no protection left, which is why RAID 6 (or erasure coding) is the usual choice for large drives.
Random-write performance
Theoretical random-write IOPS = drives × IOPS per drive ÷ write penalty. Twelve hard drives at 180 IOPS give about 2,160 read IOPS but only 360 random-write IOPS in RAID 6 — use flash or a caching tier for write-heavy workloads.
RAID is not a backup
RAID protects against drive failure, not deletion, corruption, ransomware or site loss. Pair it with backups — see the Backup & Replication Storage Calculator.
Compare levels with your own drives in the RAID & Storage Capacity Calculator.