RAID 6
RAID 6
Block-level striping with double distributed parity (P + Q). Survives any 2 simultaneous drive failures. The default for arrays with 5+ drives or drives ≥ 12 TB.
How it works
Like RAID 5 but with a second independent parity (Q, often Reed-Solomon). Both parities rotate across drives. Two drives can fail simultaneously and the array still rebuilds. Tolerates a single URE during rebuild because the second parity can fill in.
Formula: (N − 2) × min(drives)
Pros / Cons
Pros
- Survives 2 simultaneous drive failures
- Tolerates a URE during rebuild — critical for large drives
- Excellent read performance
- Mature, supported everywhere
- Good capacity efficiency vs RAID 10
Cons
- Loses 2 drives of capacity
- Write penalty (6 I/Os per small write)
- Slower writes than RAID 5
- Longer rebuild than mirrors
- Not ideal for high-IOPS workloads
When to use
Arrays with 5+ drives. Drives ≥ 12 TB. Bulk file storage, backup targets, media archives. Mission-critical pools where 2-drive failure must not lose data.
When NOT to use
High IOPS workloads (use RAID 10). Small pools (3-4 drives, ≤ 8 TB) — RAID 5 acceptable. ZFS users — use RAIDZ2 for checksums.
Rebuild math example
6 × 8 TB in RAID 6. After one drive fails: reads (6-2) × 8 = 32 TB at 70 MB/s ≈ 32 hours. URE probability (~10⁻¹⁵/bit) ~22%, but RAID 6 tolerates it — second parity reconstructs. Without that buffer, the array would likely die.