RAID Calculator

This calculator turns an array layout into the numbers that matter when you order or build a server: how much of the storage you are paying for is actually usable, how much goes to redundancy, and how many drives can fail before data loss. Choose RAID 0, 1, 5, 6, or 10, set the number of drives and their size, and it works out raw and usable capacity, efficiency, overhead, and fault tolerance.

Results are shown in both terabytes (TB, the decimal units on the drive label) and tebibytes (TiB, the binary units most operating systems report), so the 4 TB drive that shows up as 3.64 in your OS stops being a surprise.

How RAID capacity is calculated

Every RAID level is a simple multiplier on the space you buy. With n drives of size s, RAID 0 stripes across all of them and gives you the full n × s; nothing is spent on redundancy. RAID 1 mirrors, so however many drives are in the set, usable capacity is one drive: 1 × s. RAID 5 spends one drive’s worth of capacity on parity, leaving (n − 1) × s; RAID 6 spends two on dual parity, leaving (n − 2) × s. RAID 10 mirrors pairs and stripes across them, so half the drives hold copies and you keep (n ÷ 2) × s.

That is the entire calculation. In RAID 5 and 6 the parity is distributed across every member rather than parked on one disk, but the accounting still works out to exactly one or two drives’ worth. Efficiency is just usable divided by raw: a 4-drive RAID 5 runs at 75%, an 8-drive RAID 5 at 87.5%, while RAID 10 sits at 50% no matter how many pairs you add, and RAID 1 falls further the more copies you keep.

• RAID 0: usable = n × s. All capacity, no redundancy.
• RAID 1: usable = 1 × s. Every drive is a complete copy.
• RAID 5: usable = (n − 1) × s. One drive’s worth of distributed parity.
• RAID 6: usable = (n − 2) × s. Two drives’ worth of distributed parity.
• RAID 10: usable = (n ÷ 2) × s. Half the drives hold mirror copies.

Why 4 TB shows up as 3.64 TiB

Drive manufacturers count in decimal: the 4 TB on the label is 4 × 10¹² bytes, exactly as the SI prefix says. Operating systems mostly count in binary: what Windows and many Linux tools display as a "TB" is really a tebibyte (TiB): 2⁴⁰ bytes, a slightly larger unit. Dividing one by the other gives the conversion factor: 1 TB ≈ 0.9095 TiB. Nothing is missing and nobody is short-changing you; the same bytes are being measured with two different rulers.

That is why a 4 TB drive appears as roughly 3.64 TiB, an 8 TB drive as about 7.28 TiB, and a 12-drive RAID 5 of 4 TB drives (44 TB usable on paper) as around 40.02 TiB once the OS reports it, all before the filesystem takes its own small cut. The calculator shows both units side by side so the capacity you planned and the capacity the server reports line up.

RAID 0 vs 1 vs 5 vs 6 vs 10

RAID 0 stripes with no redundancy: full capacity and the combined throughput of every drive, but any single failure loses the whole volume. It is only appropriate for data you can afford to lose instantly: scratch space, caches, reproducible build artifacts. RAID 1 is its opposite: every drive is a complete copy, capacity is one drive’s worth, and the failure story is as simple as storage gets; a single surviving drive is a readable copy of everything.

RAID 5 keeps (n − 1) drives of capacity and survives any one failure, which reads like the best deal in the table. The fine print is the rebuild: replacing a failed drive means reading every byte of every surviving drive, and on large disks that window can stretch long, with zero redundancy and maximum sustained load the whole time. A second failure or an unreadable sector mid-rebuild ends the array. That risk grows with drive size, which is why single parity has aged poorly as drives have grown, and why RAID 6 exists: dual parity costs a second drive of capacity but keeps the array protected through exactly that rebuild window.

RAID 10 pays the mirror’s 50% capacity tax and buys the calmest operational profile of the group: strong random-write behaviour with no parity math, and a rebuild that is a straight copy of one mirror rather than a whole-array reconstruction. It always survives one failure and can survive up to half the drives if each failure lands in a different pair, though two failures in the same pair are fatal, which is why one guaranteed drive is the honest number to plan around.

RAID is availability, not backup

RAID answers exactly one question: does the server keep running when a drive dies? Every other data-loss mode passes straight through it: accidental deletions, application bugs, ransomware, and filesystem corruption are replicated to every drive in the array within milliseconds, and a controller failure or a second drive death can still take the whole volume. Whatever level you pick here, it complements backups; it never replaces them.

As for which level: databases, VM hosts, and anything random-write-heavy generally want RAID 10. Read-mostly bulk storage (media libraries, archives, backup targets) is where RAID 5 and 6 earn their capacity advantage, with RAID 6 the safer call on large drives. Two-drive servers take RAID 1 almost by default, and RAID 0 is reserved for data whose loss costs you a re-run rather than an incident.