A single-parity array presents the capacity of all its members but one, with the smallest member setting the size that each contributes. Three drives is the minimum the level can be built from.
The arithmetic is short. What it does not cover is longer, and more important.
The calculation itself
A single-parity array stores, alongside the data, enough parity information to reconstruct any one member that is lost. That parity is spread across all the drives rather than living on one of them, but the amount of it is equivalent to one member’s worth.
So the usable capacity is the number of drives, less one, multiplied by the size of the smallest drive in the set. Four drives of the same size present three drives’ worth. Five present four.
Double parity — the level built to survive two simultaneous failures — gives up the equivalent of two members instead, and needs at least four drives to exist at all.
Mirroring presents the capacity of one member regardless of how many copies are held.
The smallest member rule, and why it bites
Every member contributes the size of the smallest one in the set. A four-bay array holding three large drives and one small one is an array of four small drives as far as capacity is concerned, and the excess on the large ones sits unused.
This is what makes an expansion plan worth writing down before the first drive is bought. Some enclosure makers offer proprietary arrangements that recover part of that excess; each publishes its own rules, and those rules are specific to that maker rather than to the level.
Read the figure in the unit it is given
The drives are sold in decimal terms and many enclosure interfaces display in units of 1024. The array will therefore report a smaller-looking figure than multiplying the numbers on the boxes suggests, and nothing has gone wrong. Plan in whichever unit the interface shows, because that is the number you will watch fill, and compare drives against drives in the maker’s unit.
The part the arithmetic does not cover
A parity array survives the loss of one member. It does not survive a file deleted by mistake, a volume encrypted by hostile software, a theft or a fire — every one of which reaches all the members at once, because they are all in the same box.
There is a second thing worth knowing before choosing a level for capacity reasons. Replacing a failed member starts a rebuild, and a rebuild reads every remaining member from end to end while the array is still in use. It is the heaviest work the drives will ever do, it takes longer as capacity grows, and it happens at the moment the array has the least redundancy left. That is the argument for double parity on large arrays, and it is an argument about risk rather than about space.
How to decide
Write down the capacity you need, choose the level that provides the redundancy you want, and then work backwards to the drive size and the bay count — in that order. Choosing the level to maximise usable space is how people end up with a large array and a rebuild they would rather not be running.
The questions that come up before an order
Can I mix drives of different sizes in one array?
Most controllers allow it and the excess on the larger drives is not used, because every member contributes the size of the smallest. Some enclosure makers offer a proprietary arrangement that recovers part of that excess, and each publishes its own rules and its own calculator for it.
Does the usable figure appear in the same unit as the drives?
Not usually. Drives are sold in decimal terms and many enclosure interfaces display in units of 1024, so the array reports a smaller-looking figure than the arithmetic on the boxes suggests. Both are correct in their own convention, and neither is a fault.
Last reviewed 10 September 2026