RAID Calculator
Plan a NAS or server array. Enter your drives, including mixed sizes, and pick a RAID level to see the real usable space, how many drives can fail, how long a rebuild takes and how likely it is to hit a read error. Every level is compared side by side.
- RAID, SHR and ZFS
- Rebuild and read error risk
- Speed and efficiency
Result
Rebuild and read errors
Rough performance
Compared with a single drive, before network, CPU and caching limits. A 1 Gbps network tops out at about 115 MB/s.
Every level with these drives
Click a level to see its details. Read error risk is for a rebuild after one drive fails.
| Level | Usable | Efficiency | Can fail | Rebuild |
|---|
How the read error risk is worked out
Hard drive datasheets quote an unrecoverable read error (URE) rate, such as "less than 1 in 10¹⁴ bits". 10¹⁴ bits is 12.5 TB. When a drive fails, the array rebuilds it by reading every other drive in full. If that's 24 TB (three 8 TB drives), that's 1.92 × 10¹⁴ bits, so at exactly the datasheet rate you'd expect about 1.9 errors.
We work out the chance of at least one error as 1 − (1 − 1/rate)bits read, which for 24 TB at 1 in 10¹⁴ is about 85%. That's the famous "RAID 5 is dead" sum.
Real drives usually do much better. The datasheet figure is a worst-case limit the maker promises not to exceed, not a measured average. If it were typical, reading a whole 12 TB drive would fail more often than not, and that isn't what people see. So treat our figure as a pessimistic upper bound. It's still a good reason to use double parity with big drives, and to scrub regularly so bad sectors are found and fixed before a drive fails.
Which level should I pick?
- 2 drives: RAID 1 or SHR (the same thing with two drives).
- 3 to 5 drives at home: SHR or RAID 5 for small drives; SHR-2 or RAID 6 once drives are over about 4 TB.
- Business file storage: RAID 6 or RAIDZ2. Two drives can fail and rebuilds are much safer.
- Virtual machines and databases: RAID 10 or ZFS mirrors, for the best random performance and the quickest rebuilds.
- Mixed drive sizes: Synology SHR uses the extra space on bigger drives that RAID 5 and 6 would waste.
Good to know
- Drive makers use decimal terabytes (1 TB = 10¹² bytes). Windows and many NAS screens use binary units (TiB), so an 8 TB drive shows as about 7.28 "TB".
- Traditional RAID rebuilds the whole drive even if it's nearly empty. ZFS and Synology on Btrfs are smarter in places: ZFS only resilvers the space in use.
- ZFS keeps back about 3% of a pool (1/32, capped at 128 GiB) as "slop space" so it never fills completely. Most guides also suggest keeping a pool under about 80% full so it stays fast.
- RAIDZ performance for random reads and writes is roughly that of one drive per vdev, which is why big pools use several vdevs.
- RAID is not a backup. Plan proper backups with the Backup Size & Storage Planner.