Plan your RAID array before purchase. See exactly how much usable storage you get, how many drives can fail, and storage efficiency for every RAID level.
WHICH RAID LEVEL?
RAID 5 is the most common for NAS/server use — good balance of storage efficiency (~75% with 4 drives) and fault tolerance. RAID 6 survives 2 drive failures. RAID 10 has the best rebuild performance but uses 50% of capacity. Never use RAID 0 for important data.
RAID Calculator
Works out usable capacity, fault tolerance and efficiency across RAID 0, 1, 5, 6 and 10 — before you buy drives, not after you've racked them and discovered the usable space is a lot less than the sum of the labels suggested.
Which level actually fits your situation
- RAID 0 — full capacity, zero redundancy. One drive dies, everything's gone. Fine for a scratch disk or cache, never for anything you'd be upset to lose.
- RAID 1 — mirrored, 50% efficiency, survives one drive failure. Good fit for small, critical volumes like an OS drive.
- RAID 5 — striping plus parity, minimum 3 drives, survives one failure. The usual default for NAS and general server storage — good balance of capacity, performance and redundancy.
- RAID 6 — double parity, minimum 4 drives, survives two simultaneous failures. Worth the extra drive once you're using large disks (4TB+), because rebuild times stretch out and a second failure during a long rebuild is a real risk, not a theoretical one.
- RAID 10 — mirrored pairs striped together, minimum 4 drives (even number), 50% efficiency. Fastest rebuilds, best performance under load — the choice for busy databases and virtualisation hosts where rebuild time under production load actually matters.
The capacity maths
- RAID 0: drives × drive size
- RAID 1: drive size (only one copy counts toward usable space)
- RAID 5: (drives − 1) × drive size
- RAID 6: (drives − 2) × drive size
- RAID 10: (drives ÷ 2) × drive size
The warning that has to be said every time
RAID is not a backup. It protects against a drive failing, full stop — not against someone deleting the wrong folder, ransomware encrypting the array, fire, or the controller itself dying. Follow 3-2-1: three copies, two different media types, one offsite. And with large drives (8TB+), RAID 5 rebuilds can run past 24 hours, during which the array has zero redundancy — which is exactly why RAID 6 or something like ZFS is the more sensible choice once your drives get that big.