RAID Capacity Calculator
Usable capacity, fault tolerance and write penalty for every common RAID level.
How it works
Every RAID level trades three things against each other: capacity, redundancy and write speed. There is no level that is best at all three, and choosing one means deciding which you can most afford to lose.
RAID 5 gives up one disk's worth of capacity and survives one failure. RAID 6 gives up two and survives two. RAID 10 gives up half of everything and survives one failure per mirror — potentially several disks, as long as no pair goes together.
The write penalty is the part that is missing from most comparisons and matters most in practice. A single small write to RAID 5 requires reading the old data, reading the old parity, writing the new data and writing the new parity — four operations for one logical write. RAID 6 needs six. RAID 10 needs two. On spinning disks that difference is the whole reason databases run on RAID 10 despite paying half their capacity for it.
Then there is rebuild. When a disk in a RAID 5 array fails, every remaining disk must be read end to end to reconstruct it, and with modern high-capacity drives that takes many hours to days at full load. During that window there is no redundancy at all, and the read stress is precisely when a second marginal drive tends to give up. This is why RAID 5 on large drives is widely considered obsolete — the array is most likely to fail exactly when it is least able to survive it.
The unit confusion is worth clearing up: a "4 TB" disk is 4 × 10¹² bytes, which the operating system reports as 3.64 TiB. The missing 9% is not lost, it is two different definitions of the same prefix.
None of this is a backup. RAID protects against a disk dying. It does not protect against deletion, corruption, ransomware, a controller writing garbage to every member at once, or the building burning down — all of which it faithfully replicates across every disk in the array.
Common questions
RAID 5 or RAID 6?
RAID 6 for anything with drives above about 2 TB. Rebuild takes long enough on large drives that a second failure during it stops being unlikely, and RAID 5 has no redundancy left at that moment. The extra disk is cheap insurance.
What is the write penalty?
The number of physical I/O operations one logical write costs. RAID 0 and JBOD 1, RAID 1 and 10 2, RAID 5 4, RAID 6 6. It only bites on small random writes; large sequential writes fill whole stripes and avoid the read-modify-write cycle.
Why is my array smaller than advertised?
Two reasons stacked. Drives are sold in decimal TB and reported in binary TiB, which loses about 9%, and then the filesystem takes a few percent for metadata. A 6 × 4 TB RAID 5 gives 20 TB raw usable, about 18.2 TiB, and roughly 17.6 TiB after formatting.
Is RAID a backup?
No, and treating it as one is the most expensive mistake in storage. RAID survives a disk failing. It faithfully replicates a deletion, a corruption or an encryption to every disk in the array, instantly. You still need backups that are somewhere else.