RAID calculator

Pick a RAID level and the calculator shows what you actually get: usable capacity, how many disks can fail before the array is gone, and roughly how fast reads and writes will be compared with one disk on its own.

Capacity appears twice on purpose. Disks are sold in decimal terabytes — a "4 TB" disk holds 4,000,000,000,000 bytes — while Linux, macOS and Windows report tebibytes, which are 10% larger each. That gap is not lost space and it is not a faulty disk; it is two different units with confusingly similar names.

How usable capacity is worked out

Every level sets aside some disks for redundancy. What is left is the capacity you can write to, and it is always a whole number of disks multiplied by the size of the smallest one — mixing a 4 TB disk into an array of 2 TB disks gives you 2 TB of it.

• RAID 0 — all disks carry data. Usable = disks x size. Nothing is protected: one failure loses everything.
• RAID 1 — every disk holds the same data. Usable = one disk, and all but one may fail.
• RAID 5 — one disk's worth of parity. Usable = (disks - 1) x size, one failure survived.
• RAID 6 — two disks' worth of parity. Usable = (disks - 2) x size, two failures survived.
• RAID 10 — mirrored pairs, then striped. Usable = half. One failure is survived for certain, more only if no two land in the same pair.
• RAID 50 and 60 — several RAID 5 or 6 groups striped together. Each group pays its own parity, so more groups means less capacity and faster rebuilds.

The write penalty nobody quotes

Sequential throughput is the number most calculators stop at, and it is the one that matters least for a busy server. Parity RAID has to read the old data and the old parity before it can write new ones, so a single small write costs four physical I/Os on RAID 5 and six on RAID 6. RAID 10 costs two.

That penalty is why a database on RAID 6 can feel slow on disks that benchmark beautifully. Divide the array's raw IOPS by the penalty to get what the application will actually see, and size from that number.

Rebuild time is the real risk

An array is at its most fragile while it rebuilds after a failure: every remaining disk is read end to end, which is exactly the workload most likely to expose a second bad disk. With today's large drives a rebuild runs for many hours, sometimes days.

That is the argument for RAID 6 or RAID 10 over RAID 5 on anything large, and the reason RAID is not a backup. RAID keeps a service running through a disk failure. It does nothing about a deleted file, a bad deploy or ransomware — only a separate copy does.

Common questions

Why is my usable capacity smaller than the calculator says?

Three common reasons: the filesystem reserves space (ext4 keeps 5% for root by default), the controller keeps metadata, and your OS reports tebibytes rather than the decimal terabytes the disk was sold in. The calculator shows both units so you can tell which one you are looking at.

Is RAID 5 safe to use?

On small arrays of small disks, yes. On large modern disks the rebuild window is long enough that a second failure during it is a real risk, so RAID 6 or RAID 10 is the safer choice above about 4 disks.

RAID 10 or RAID 6?

RAID 10 for write-heavy work such as databases: half the capacity, but the write penalty is 2 instead of 6 and rebuilds only read one disk. RAID 6 when capacity matters more than write speed and you want two disks of protection across the whole array.

Does RAID replace backups?

No. RAID protects against a disk dying. It copies a deletion, a corruption and a ransomware run to every disk just as faithfully as it copies your data. Keep a separate backup somewhere the server cannot reach.

Can I use disks of different sizes?

Usually yes, but every disk is treated as if it were the size of the smallest one, so the extra capacity is wasted. Some systems (Synology SHR, ZFS with care, unRAID) can use part of it; standard hardware RAID cannot.