How to choose a RAID level: matching redundancy to how much you can lose

Once you understand why a RAID array reports less space than the drives promise, the next question is the one that actually costs money: which RAID level should you build? The levels aren't ranked from worst to best — they're a menu of trade-offs between how much usable space you keep, how many drives can die before you lose everything, how long and how risky the rebuild is when one does, and how fast the array runs day to day. Picking well means being honest about which of those you can least afford to compromise.

What each level actually costs you

Start with the capacity math, because it's the most visible trade. With n drives of equal size c: RAID 0 gives you the full n × c and tolerates zero failures — one dead drive loses the whole array. RAID 1 mirrors, so a pair gives you c and survives one failure; usable space is roughly half. RAID 5 gives you (n − 1) × c and survives one failure, spending exactly one drive's worth on parity. RAID 6 gives you (n − 2) × c and survives two failures. RAID 10 mirrors then stripes, giving you (n ÷ 2) × c and surviving at least one failure — potentially more, as long as you don't lose both halves of the same mirror.

So RAID 0 is pure capacity with no safety net, mirroring and RAID 10 spend half your drives on redundancy, and the parity levels (5 and 6) sit in between, giving up one or two drives regardless of how many you have. That last point matters: in a large array, RAID 6's two-drive overhead is a small percentage; in a four-drive array, it's fully half.

The rebuild is where arrays actually die

Fault tolerance on paper is not the same as safety in practice, and the gap is the rebuild. When a drive fails in a single-parity array like RAID 5, the array is degraded until you replace the drive and it rebuilds — and during that rebuild, every remaining drive is read in full, under load, often for many hours on large modern disks. Two things make this dangerous. First, a second failure during the rebuild loses everything, and drives bought together tend to age and fail together. Second, even without an outright failure, an unrecoverable read error on any surviving drive can abort the rebuild. As drive capacities climbed into the 8, 12, and 16+ TB range, the odds of hitting a read error while reading an entire drive stopped being negligible.

This is the single biggest reason the storage community pushes RAID 6 over RAID 5 for large drives: the second parity drive means a rebuild can survive both a read error and a concurrent failure. If you're building an array out of big disks — say, anything from 8 TB up — the one-drive cushion of RAID 5 is thinner than the capacity math suggests.

A decision path

Choose RAID 0 only for scratch space you can lose without a second thought — video editing caches, temporary renders, anything already backed up elsewhere and valued purely for speed. Never for the only copy of anything.

Choose RAID 1 for small setups, typically two drives, where simplicity matters and you don't need much space. Rebuilds are simple copies, performance is solid for reads, and there's no parity math to go wrong. It's a fine choice for a two-bay NAS.

Choose RAID 5 for four to six smaller drives where you want more usable space than mirroring allows and the individual drives aren't huge. It's the efficient middle ground — but weigh the rebuild risk as your drive size grows.

Choose RAID 6 for larger arrays and larger drives, where the survivability of a two-drive failure window during rebuild is worth giving up a second drive's capacity. For most people building a serious multi-bay NAS with high-capacity disks today, this is the default.

Choose RAID 10 when performance under mixed read/write load matters — databases, busy virtualization — and you're willing to spend half your capacity for it. Rebuilds are fast because they're straight mirror copies rather than parity recalculation, and write performance beats the parity levels.

The rule RAID doesn't change

Whatever level you pick, none of them is a backup. RAID protects against a dead drive — it keeps the array online while you replace hardware. It does nothing about a deleted file, a ransomware sweep, a controller that corrupts data as it writes, or a house fire that takes the whole box. Those are exactly the failures that erase real data, and the only defense is a genuine backup on separate media, ideally with a copy off-site. Decide your RAID level for uptime and capacity; decide your backup strategy for actually not losing anything.

To see the trade laid out in numbers for your specific plan, set your drive count, capacity, and level in the NAS / RAID capacity calculator — it shows usable space after overhead and how many drives can fail alongside it, so you can compare RAID 5, 6, and 10 for the same hardware before you buy.