RAID Calculator

RAID Calculator

Choose a RAID level, enter the number of disks and the capacity of each one (in GB or TB), and the calculator shows:

• Usable capacity, raw capacity and how much space goes to redundancy
• Storage efficiency
• How many drives can fail without losing data
• A rough performance estimate

For RAID 50 and 60 there's an extra field for the number of RAID groups, explained below. Everything is calculated in your browser.

Example: four 4 TB disks in RAID 5

Usable capacity. One disk's worth of space goes to parity, the extra data that lets the array rebuild a failed drive. The parity isn't stored on one dedicated disk; it's spread across all four, so no single drive becomes a bottleneck.

Efficiency. Add a fifth disk and efficiency rises to 80% (16 TB usable out of 20 TB), because the parity cost stays at one disk however many you add. The catch is that more disks also means more chances of a failure, and RAID 5 still only covers one.

Performance. The "~3× read, ~3× write" estimate applies to large sequential transfers, such as copying big files. Small random writes behave very differently on RAID 5. Each one costs four disk operations: read the old data, read the old parity, write the new data, write the new parity. For databases or virtual machines, a four-disk RAID 5 writes at roughly the speed of a single disk.

What your computer will show. Drive makers count a terabyte as 1,000,000,000,000 bytes. Windows and many NAS systems count in binary units, where a "TB" is about 10% larger. So these 12 TB will show up as roughly 10.9 TB once formatted. The Real Disk Space Calculator works this out for any drive size.

Same four disks, different levels

RAID 6 and RAID 10 give exactly the same capacity with four disks, so the choice comes down to what matters more:

• RAID 6 survives any two failures.
• RAID 10 is faster and rebuilds much quicker. But if a second disk fails before the first is replaced, there's a one-in-three chance it's the partner of the dead one, and then the whole array is gone.

The levels

RAID 0: striping

Data is split across all the disks with no redundancy. You get the full capacity and the best speed, and if any one disk fails, everything on the array is lost. Each disk you add makes that more likely. Use it for scratch space, render caches and anything you can recreate. Never use it for anything you can't.

RAID 1: mirroring

Every disk holds a full copy of the data. Reads can come from any disk; writes go to all of them, so writing is no faster than with one drive. It's the usual choice for boot drives and two-bay NAS boxes. Rebuilding is just copying one disk to another.

RAID 5: single parity

RAID 5 needs at least three disks and gives you all but one disk of capacity. Its weak point is the rebuild. When a drive fails, the array has no protection until the replacement is fully rebuilt, and rebuilding means reading every sector of every remaining disk.

Large drives make that window long and risky. A rebuild of 8 TB drives takes at least half a day even if nothing else is using the array, and often days in practice. Consumer and many NAS drives are rated for one unrecoverable read error per 10^14 bits, which is about 12.5 TB of reading. Real drives usually do better than their rating, but a single bad read during a rebuild can be enough to fail it. This is why RAID 5 has fallen out of favor for arrays of big disks.

RAID 6: double parity

RAID 6 is like RAID 5 with a second, independent parity block. It needs at least four disks and gives up two disks of capacity. In return, it survives two failures, or one failure plus a read error during the rebuild. That second scenario is the main reason to choose it.

Small random writes are slower than on RAID 5: six disk operations each instead of four.

RAID 10: mirrors, striped

Disks are paired into mirrors, and the mirrors are striped together. It needs an even number of disks, at least four, and you always get exactly half the raw capacity. Random writes cost only two operations, and a rebuild just copies the surviving partner, so it's fast. That's why databases and busy virtual-machine storage usually run on RAID 10.

RAID 50 and RAID 60: groups

These split the disks into several RAID 5 (or RAID 6) groups and stripe across them. The Number of RAID Groups field sets how many groups there are, and the disks have to divide evenly between them. For example, with twelve 4 TB disks:

• RAID 50 in two groups of six: 40 TB usable, survives one failure in each group.
• RAID 60 in two groups of six: 32 TB usable, survives two failures in each group.

Each group rebuilds on its own, which is faster and less risky than one wide RAID 5 or 6. The trade-off is that losing too many disks in the same group still takes down the whole array.

Before you buy the disks

Use disks of the same size. The calculator assumes equal disks, and that's how traditional RAID works: every disk is treated as the size of the smallest one. Synology's SHR and similar systems handle mixed sizes differently, so these numbers won't match them.

Don't count a hot spare. A hot spare sits idle until a disk fails, then the rebuild starts on it automatically. It adds no capacity, so leave it out of the disk count.

Avoid SMR drives. Shingled (SMR) drives can make rebuilds extremely slow or drop out of the array entirely. Drives sold for NAS use are mostly CMR, but check the spec sheet.

Mix your batches. Drives from the same production batch, bought on the same day, tend to wear out around the same time. That's exactly what you don't want during a rebuild.

Using ZFS? For capacity planning, RAID-Z1 behaves roughly like RAID 5, RAID-Z2 like RAID 6, and striped mirrors like RAID 10. ZFS keeps some space in reserve, so expect a little less.

RAID is not a backup. It keeps the array running when a disk dies. It does nothing about deleted files, ransomware, a failed controller or a fire. Keep a separate copy somewhere else.

FAQ

Why does my NAS show less space than the calculator?

Mostly because of the decimal vs. binary terabyte difference explained above, which accounts for about 9%. The filesystem and any snapshot reserve take a little more.

Is RAID 5 still safe to use?

With smaller drives and good backups, it's a reasonable choice. Once drives reach around 8 TB or more, most admins choose RAID 6, because rebuilds take so long that a second problem during the rebuild becomes a real risk.

Can I add disks later?

It depends on your NAS or RAID controller. Many can expand a RAID 5 or 6 array one disk at a time, but the reshape can take days. RAID 10 normally grows by adding disks in pairs. Check what your system supports before you pick a level.