UK’s trusted IT infrastructure partner since 2003
Servnet
ConfiguratorGet in Touch

RAID 10 calculator

Striped mirrors — top write performance and fast rebuilds. Set your drives below for live usable capacity, fault tolerance, IOPS, rebuild time and URE risk.

DataDistributed parity

1 · Choose a RAID level

Stripe & mirror
Single parity
Dual / triple parity
Nested
ZFS RAID-Z

Striped mirrors. Guaranteed one failure; up to one per mirror pair if losses do not collide.

2 · Configure drives

3 · Drive class

12G SAS SSD — indicative figures.

Advanced — read/write mix, URE rate
RAID 10 · 8 × 8 TB
32 TB usable
of 64 TB raw · 50% efficiency
Fault tolerance1 guaranteed; up to 4 (one per mirror pair) if losses don't collide
Write penalty×2
IOPS estR ≈600K · W ≈300K · mix ≈462K
Throughput estR ≈8K · W ≈4K MB/s
Rebuild / drive est≈ 4.4 h
URE on rebuild risk0.71%

During a single-drive rebuild there is no remaining redundancy — a URE on a surviving drive means data loss for the affected stripe. Real controllers mitigate via patrol reads/scrubs, so field results are often better.

Capacity distribution50% usableUsable: 32 TB32Parity: 32 TB32Usable · 32 TBParity · 32 TB
Fault tolerance — parity per groupgroup 1DPgroup 2DPgroup 3DPgroup 4DPDataParity1 guaranteed; up to 4 (one per mirror pair) if losses don't collide
IOPS — back-end budget vs deliveredBack-end budget600KFront-end read600KFront-end write300KWrite penalty ×2 — each host write costs 2 back-end I/Os

Calculated for planning. We don't publish prices — a 24-year UK reseller, Servnet confirms the exact drives, array and pricing on quote. IOPS, throughput & rebuild are indicative estimates.

Overview

What RAID 10 is

RAID 10 stripes data across mirrored pairs, combining RAID 1 redundancy with RAID 0 striping. Usable capacity is half the raw (n/2 × drive size) and it delivers the best write performance of the resilient levels thanks to a ×2 write penalty.

It always survives one failure and can survive up to one drive per mirror pair if losses do not collide. Rebuilds are fast — just a copy from the surviving mirror, not a parity recalculation across the whole array — which is why RAID 10 is the standard for write-heavy, latency-sensitive databases despite its 50% efficiency.

At a glance
Usable capacityn / 2 × drive size (50%)
Minimum drives4 (even)
Fault tolerance1 guaranteed; up to 1 per pair
Write penalty×2
Worked example
8 × 8 TB SAS SSD32 TB usable, ×2 write penalty

Eight 8 TB SSDs in RAID 10 give 32 TB usable and the highest write IOPS of any resilient level — each host write costs only two back-end I/Os. The price is 50% efficiency: 64 TB raw for 32 TB usable.

Advantages

  • Best write performance of the resilient levels (×2 penalty)
  • Fast, low-impact rebuilds (mirror copy)
  • Excellent random I/O for databases / VMs
  • No parity-calculation overhead

Trade-offs

  • Only 50% capacity efficiency
  • Guaranteed to survive just one failure (more only if lucky)
  • Expensive per usable TB
  • Needs an even number of drives

Best for

  • Write-heavy transactional databases (SQL, Oracle)
  • Latency-sensitive virtualisation / VDI
  • Mixed random I/O workloads
  • Anywhere rebuild speed is critical

Consider another level when

  • Capacity-led bulk / archive storage
  • Budget-constrained capacity needs
  • Read-mostly workloads where RAID 6 is cheaper
Level landscape — efficiency vs fault tolerance (typical)012325%50%75%100%drives survivedspace efficiency →RAID 0RAID 5RAID 50RAID-Z1RAID 6RAID 60RAID-Z2RAID-Z3RAID 10RAID 1

RAID 10 — common questions

How much usable capacity does RAID 10 give?

Half the raw capacity — n/2 × drive size. Eight 8 TB drives give 32 TB usable (50% efficiency), because every drive is mirrored.

How many drives can RAID 10 lose?

It always survives one failure. It can survive more — up to one drive per mirror pair — but only if no two failures land in the same pair. The guaranteed (worst-case) figure is one drive, which is what the calculator reports.

Why is RAID 10 better for databases?

Its ×2 write penalty (versus ×4 for RAID 5 and ×6 for RAID 6) means far higher write IOPS, and rebuilds are a fast mirror copy rather than an array-wide parity recalculation — both of which matter for write-heavy, latency-sensitive databases.