RAID 10

Striped mirrors. Each pair is RAID 1, then RAID 0 across pairs. The fastest traditional RAID for IOPS workloads — at the cost of half the raw capacity.

Min. drives
4
Usable capacity
N/2 × pair
Fault tolerance
1+ per mirror pair
Performance
Best IOPS in classical RAID

How it works

Drives are grouped into mirror pairs (RAID 1). Data is striped across the pairs (RAID 0). Each block exists on exactly two drives. A drive can fail in any pair without data loss. Rebuilds are fast because only the surviving partner is read.

Formula: (N / 2) × min(pair)

RAID 10 — RAID 1 within pairs, RAID 0 acrossMirror pair AMirror pair BD1A1A2A3D2A1A2A3D3B1B2B3D4B1B2B3
Layout diagram

Pros / Cons

Pros

  • Excellent random read AND write IOPS
  • Survives 1+ failures per mirror pair (potentially many across the array)
  • Fast rebuilds — only copies the surviving partner
  • Simple mental model
  • No parity write penalty

Cons

  • Halves total capacity (50% efficient)
  • Even drive count required
  • Capacity-inefficient vs RAID 6
  • Pair-coupled failure: both drives in one pair = data loss

When to use

VMs, busy databases, high-IOPS workloads. Production app servers. Anywhere write IOPS matters more than capacity.

When NOT to use

Bulk storage / archives — RAID 6 or RAIDZ2 buy you safety with less overhead. Tiny budgets — too much capacity overhead.

Rebuild math example

8 × 8 TB in RAID 10 (4 pairs). After one drive fails: reads its partner (8 TB) at 70 MB/s ≈ 8 hours. URE probability stays low because only one drive is read.

Check rebuild risk →

Try it now

Related

Bottom line: RAID 10 trades capacity for IOPS and rebuild speed. Pick it when write performance and quick recovery matter more than raw TB.