Free RAID calculator for RAID 0, 1, 5, 6, 10, 50 and 60. See usable capacity, what redundancy costs, and how many disks can really fail.
What each level costs you before you enter any disks
| Level | Min disks | Redundancy |
|---|---|---|
| RAID 0 | 2 | None |
| RAID 1 | 2 | All but one disk |
| RAID 5 | 3 | One disk |
| RAID 6 | 4 | Two disks |
| RAID 10 | 4 | Half the disks |
| RAID 50 | 6 | One disk per group |
| RAID 60 | 8 | Two disks per group |
Redundancy is what each level spends before you store anything. Enter a disk count and a size to see what all seven give you with your own disks.
A disk sold as 1 TB holds 1,000,000,000,000 bytes; the OS divides by 2⁴⁰
| On the box | Your OS shows |
|---|---|
| 1 TB | 0.91 TiB |
| 2 TB | 1.82 TiB |
| 4 TB | 3.64 TiB |
| 8 TB | 7.28 TiB |
| 12 TB | 10.91 TiB |
| 16 TB | 14.55 TiB |
| 20 TB | 18.19 TiB |
Drive makers count in powers of ten and operating systems count in powers of two, so the same disk reads about 9.05% smaller once it is formatted. Nothing is missing — TB and TiB are two different units, and NIST names the second set kibi, mebi, gibi and tebi for exactly this reason. Formatting and filesystem metadata then take a further slice on top.
Figures are the capacity arithmetic for standard RAID levels with disks of equal size. Controllers, filesystems and vendor schemes reserve space of their own, so a finished array will show slightly less than the usable figure above.
You might also find these calculators useful
Pick a level, say how many disks you have and how big they are, and this calculator returns the three numbers that decide the build: how much you can store, how much of what you bought is spent on redundancy, and how many disks can fail before the data is gone. It also shows the capacity twice — once in the units the disks were sold in, and once in the units your operating system will report.
Every RAID level is a bargain between capacity, speed and survival. Striping alone (RAID 0) keeps all the capacity and none of the protection. Mirroring (RAID 1) keeps one disk's worth however many you add. Parity levels give up one disk of capacity (RAID 5) or two (RAID 6) and can rebuild what a failed disk held from what is left. The nested levels — 10, 50 and 60 — split the disks into groups, run one of those schemes inside each group, and stripe across the groups, which is why their fault tolerance depends on where the failures land.
RAID 5 usable capacity
4 × 4 TB gives 12 TB usable at 75% efficiency and survives one disk. The OS will report 10.91 TiB.
8 × 12 TB gives 72 TB usable, still 75% efficiency, and survives any two disks — the usual choice once a rebuild takes days.
8 × 2 TB gives 8 TB usable at 50%. One failure is always survivable, four only if they spread across the four pairs.
12 × 16 TB in two groups of six gives 128 TB usable at 66.7%, and every group tolerates two failures of its own.
An eight-disk RAID 10 is four mirrored pairs. It survives one failure wherever it lands, and four only if no two hit the same pair. Calculators that print "4" are quoting the best case as though it were guaranteed — after the first disk dies, one of the seven survivors is its mirror, and losing that one takes the array.
Four 8 TB disks in RAID 5 give 24 TB usable, and your file manager will say 21.83 TiB. Both are correct: disks are sold in powers of ten and reported in powers of two. Showing one without the other is where "my array is missing 2 TB" comes from.
The comparison table scores all seven levels against the disks you entered, so choosing between RAID 5, RAID 6 and RAID 10 is one read rather than three runs — including which levels your disk count cannot form at all.
Nothing is missing. Drive makers sell capacity in decimal units — a 8 TB disk holds 8,000,000,000,000 bytes — while operating systems divide by 1,024 three or four times over. The same quantity is 24 TB decimal and 21.83 TiB binary, a gap of 9.05%. NIST defines the binary names (kibibyte, mebibyte, gibibyte, tebibyte) precisely so the two can be told apart. Formatting overhead then removes a little more.
One, guaranteed. RAID 10 stripes across mirrored pairs, and the array survives as long as no pair loses both its disks. In an eight-disk array you may survive four failures, but only if each lands in a different pair — and after the first failure, one of the seven remaining disks is its mirror, so roughly 14.3% of second failures end the array. Plan around one.
RAID 6 costs one more disk of capacity and buys protection during the rebuild, which is when arrays actually die. Replacing a failed disk in RAID 5 means reading every remaining disk from end to end for hours or days; a second failure in that window takes everything. The bigger the disks, the longer that window, which is why RAID 6 is the common recommendation past about 4 TB per disk.
No. RAID protects against a disk failing. It does not protect against deleting the wrong files, ransomware, a controller fault, a power surge or a fire, because all of those reach every disk in the array at once — and the array copies the damage faithfully. RAID keeps a machine running while you swap hardware; a backup is a separate copy somewhere else.
Standard RAID truncates every disk to the size of the smallest, so a 4 TB disk beside three 8 TB disks makes it a four-disk array of 4 TB disks and the extra 12 TB is unusable. Vendor schemes such as Synology's SHR and Drobo's BeyondRAID pool mismatched disks differently by layering several arrays on the same drives; the figures here are for standard levels with equal disks.
SNIA's Common RAID Disk Data Format specification defines the primary levels — RAID 0, 1, 3, 4, 5 and 6 — and a secondary level laid across groups, which is what RAID 10, 50 and 60 are. The original 1988 Berkeley paper by Patterson, Gibson and Katz introduced levels 1 through 5; RAID 0, RAID 6 and every nested level came later as industry convention rather than from that paper.