Skip to content
ToolBoxGeniehome

RAID Capacity Calculator

Developer Tools · Added

How much space an array actually gives you, how many disks it can lose, and how much data has to be read back to rebuild after one fails. Every capacity is shown twice — as decimal terabytes, which is what the disk was sold as, and as binary tebibytes, which is what your operating system will report — because the gap between those two is the most common reason people open a calculator like this.

Parity distributed across all disks costs exactly one disk of capacity. Survives one failure. During a rebuild it survives none, and the rebuild reads every sector of every remaining disk.

Including any hot spares. RAID 5 (single parity) needs at least 3 in the array.

Idle disks that hold no data until a rebuild starts.

With mixed sizes, enter the smallest — that is what every member is treated as.

%

Optional. Metadata, journals and reserved blocks — typically low single digits.

How to use the raid capacity calculator

  1. 1Pick the RAID level. The hint under the selector says what it costs and what it survives.
  2. 2Enter the disk count and the size of each. With mixed sizes, enter the smallest — that is what every member gets treated as.
  3. 3Add hot spares if you have them, and a sub-array count for RAID 50 or 60.
  4. 4Read the usable capacity in both units, and the rebuild figure underneath it.

Examples

Six 4 TB disks in RAID 5

Input
6 disks, 4 TB each, single parity
Result
20 TB usable — 18.2 TiB as the operating system will show it

One disk of capacity goes to parity however many disks are in the array.

Where the missing capacity went

Input
A 4 TB disk
Result
3.64 TiB, and nothing has been lost

Four trillion bytes is what SI prefixes mean. Dividing by 1024 three times gives tebibytes, which share a name and are not the same size.

What a rebuild costs

Input
An eight-disk RAID 5 of 8 TB drives
Result
56 TB has to be read off the survivors to rebuild one disk

During which the array has no redundancy left, which is the whole argument for RAID 6.

About the raid capacity calculator

What the levels actually cost

Parity costs disks, not a percentage, and that is the fact the capacity figure obscures. RAID 5 loses exactly one disk's worth however wide the array, and RAID 6 loses two. So a four-disk RAID 5 is 75% efficient and a twelve-disk RAID 5 is 92% — the proportion improves as you add members, which makes wide arrays look attractive on a spreadsheet.

The risk moves the other way at the same time. A twelve-disk RAID 5 has eleven disks that all have to survive a rebuild that reads every one of them end to end. The wider the array, the better the efficiency and the worse the bet, and the two curves cross well before most people expect. This is the reasoning behind RAID 50 and RAID 60: several narrow parity sets striped together, so a rebuild touches only the disks in its own set.

Mirroring costs half the capacity flat, and buys back rebuild speed. Rebuilding a RAID 10 member copies one disk to one disk, rather than reading the whole array and recomputing parity — far faster, far less stressful on the survivors, and the reason RAID 10 keeps being chosen for databases despite the capacity.

Mixed disk sizes, and why the smallest wins

Standard RAID levels stripe fixed-size chunks across every member, so every member has to contribute the same amount. Put a 4 TB disk in an array of 8 TB disks and only 4 TB of each is usable; the rest is stranded. This calculator asks for the smallest disk size for that reason.

Vendor-specific layouts get around it. Synology's SHR and unRAID's parity scheme both pool disks of different sizes and recover most of the stranded space, at the cost of being tied to that vendor's implementation. They are not modelled here, because the arithmetic is specific to each product and a generic figure would be wrong in a way you could not check.

The number that is not on the box

The figure worth taking from a capacity calculation is often not the capacity at all — it is how much data has to be read back to rebuild one disk. That number decides how long the array spends without redundancy, and therefore how much the whole arrangement is actually protecting you.

Two arrays with identical usable capacity can differ by an order of magnitude here. A RAID 10 rebuild reads one disk. A wide RAID 5 rebuild reads everything else. If the rebuild figure this tool reports would take days on your hardware, that is the argument for a different level, or for more spares, or for accepting that the backups are doing the real work.

Frequently asked questions

Why does my 4 TB disk show as 3.64 TB?
Because two different units are sharing a name. The manufacturer means 4,000,000,000,000 bytes, which is what the SI prefix tera- means. Windows and most tools divide by 1024 three times and print the result while still calling it TB, when what they have computed is tebibytes. No capacity has gone anywhere and nothing is being hidden — the same bytes are being written in two units. The gap is about 10% at the terabyte scale and grows with every prefix.
Why is RAID 5 advised against on large disks?
Because rebuilding after a failure reads every sector of every surviving disk, and on multi-terabyte drives that takes many hours to days. Throughout that window the array has no redundancy left, so a second failure loses everything — and a rebuild is exactly the workload most likely to provoke one, since it hammers disks of the same age and batch. A single unreadable sector on any survivor can be enough. RAID 6 exists specifically to keep one parity in hand during the rebuild.
How many disks can RAID 10 lose?
One from each mirrored pair, so potentially half the array — but only one with certainty. Which disks fail decides it: on an eight-disk RAID 10 you can lose four and keep everything, or lose two and lose the array, depending on whether they were partners. That is why this tool reports a range rather than a single figure for RAID 10 and for the nested levels, where fault tolerance is per sub-array rather than per array.
Should I use a hot spare?
It shortens the window that matters. Without one, an array stays degraded until a person notices and physically swaps a disk, which on an unmonitored machine can be weeks. A hot spare starts the rebuild within minutes of the failure, entirely unattended. The cost is a disk of capacity that holds nothing until it is needed — which on a small array is a large fraction of the total, and is why spares are more common on large arrays than on four-bay boxes.
Is RAID a backup?
No, and treating it as one is how data gets lost. RAID protects against a disk failing and against nothing else. It does not protect against accidental deletion, ransomware, a filesystem corrupting itself, a controller writing garbage to every member at once, theft, fire or flood — in every one of those the redundancy faithfully replicates the problem. RAID is an availability measure: it keeps a machine running through a disk failure. Backups are a separate thing, and they need to be somewhere else.

Related tools