RAID capacity formulas
N is the number of drives, S the size of one drive, and G the number of groups in RAID 50 or 60.
| Level | Usable space | Drives that can fail | Min. drives |
|---|---|---|---|
| RAID 0 | N × S | 0 | 2 |
| RAID 1 | S | N − 1 | 2 |
| RAID 5 | (N − 1) × S | 1 | 3 |
| RAID 6 | (N − 2) × S | 2 | 4 |
| RAID 10 | N ÷ 2 × S | 1 for sure, up to N ÷ 2 if each is in a different mirror pair | 4 (even) |
| RAID 50 | (N − G) × S | 1 for sure, up to G (one per group) | 6 |
| RAID 60 | (N − 2G) × S | 2 for sure, up to 2G (two per group) | 8 |
| JBOD | N × S | 0 (data on a failed drive is lost) | 1 |
TiB = TB × 10¹² ÷ 2⁴⁰ = TB × 0.9095
Worked example
Four 8 TB drives hold 32 TB raw. In RAID 5, one drive's worth goes to parity: (4 − 1) × 8 = 24 TB usable, 75% efficient, and any one drive can fail. Your computer will likely show that as 24 × 10¹² ÷ 2⁴⁰ = 21.83 TiB. The same four drives give 16 TB in RAID 6 or RAID 10, and 32 TB in RAID 0 with no protection.
Speed: rough rules of thumb
The speed numbers in the result are rough, best-case rules of thumb for large sequential transfers, as a multiple of one drive. Real arrays are usually slower, because of the controller, the interface, caching and the mix of small random reads and writes.
- Reads can come from every drive that holds data, so striping speeds them up.
- Mirrors must write every copy, so RAID 1 writes about as fast as one drive.
- Parity levels must also update parity. A small random write takes about 4 disk operations in RAID 5 and 6 in RAID 6 (the "write penalty"), compared with 2 for mirrors.
TB vs TiB
Drive makers use decimal units: 1 TB = 1,000,000,000,000 bytes. Many operating systems count in binary units: 1 TiB = 2⁴⁰ = 1,099,511,627,776 bytes, sometimes still labeled "TB". So an 8 TB drive shows as about 7.28 TiB. No space is missing. See the data storage converter for every unit.