RAID calculator: capacity, tolerated failures and rebuild risk

How much space is left after parity, how many disks can fail, and what chance does your array have of rebuilding without a read error? Pick a RAID level and the calculation follows, with its formulas under the result.

The calculation runs in your browser: nothing you type is sent anywhere.

Capacity and risk of a RAID array

Your parameters

At least 3 for this level.

TB

In TB, as printed on the label (1 TB = 10¹² bytes).

Order of magnitude, replace it: about 80 for a 7,200 rpm disk, tens of thousands for an SSD.

MB/s

Example value, replace it: it depends on the controller and the load during the rebuild.

Datasheet value: a maximum guaranteed by the manufacturer, so a pessimistic bound.

Result

Usable capacity

24 TB

21.83 TiB

Efficiency

75%

32 TB raw across 4 disks

Tolerated failures

1 disk

Random IOPS

320 read · 80 write

Write penalty: ×4

Rebuild time for one disk

22 h 13 min

A whole disk rewritten at the given speed, without production load.

Rebuild without a read error

14.7%

Probability of reading 24 TB without a URE, once redundancy is used up.

With the datasheet value, less than an even chance of rebuilding without a read error. It is a worst case, but enough to prefer RAID 6 with large disks.

Pessimistic bound: 10¹⁴ or 10¹⁵ bits is the maximum the manufacturer guarantees, not an observed rate. Rates measured in the field are much better, and a controller often retries a sector before reporting the error. Read this probability as a floor, not as the real chance of a successful rebuild.

How the calculation works

Usable capacity depends on the RAID level, the number of disks n, their size T and, for nested levels, the number of groups g. The calculation assumes identical disks: in a mixed array, each disk counts as the smallest one.

RAID 0  : n × T                RAID 1  : T
RAID 5  : (n − 1) × T          RAID 6  : (n − 2) × T
RAID 10 : n/2 × T              RAID 50 : (n − g) × T
RAID 60 : (n − 2g) × T

Read IOPS add up across all disks. For random writes, each level pays a penalty: one write costs 1 disk operation in RAID 0, 2 in RAID 1 and 10, 4 in RAID 5 and 50, 6 in RAID 6 and 60.

read IOPS  = n × disk_IOPS
write IOPS = n × disk_IOPS / penalty

Rebuild time is that of a whole disk rewritten at the given speed. Rebuild risk is calculated when the array has no redundancy left during the operation: RAID 5 and 50 after one failure, RAID 10 and two-disk RAID 1. Every bit read can then raise an unrecoverable read error (URE) that nothing corrects:

bits_read = disks to read × T × 8
P(rebuild without error) = exp(− bits_read / 10^n)
n = 14 (consumer disk) or 15 (enterprise disk)

Sources: Standard RAID levels (Wikipedia), disk manufacturers' datasheets for URE rates.

A datasheet URE rate is a worst-case bound: real disks often do better, and a controller may retry a sector before giving up. The result shows an order of magnitude of the risk, not a prediction. IOPS per disk and rebuild speed are example values: replace them with your hardware's.

RAID is not a backup

RAID keeps a server running when a disk fails. It protects neither against a deletion, nor ransomware encrypting files, nor losing the enclosure: every mistake is written to all disks at once. A NAS in RAID 5 or 6 therefore needs a copy outside its walls. Our guide to offsite NAS backup compares the methods, with dedicated pages for offsite Synology backup and offsite QNAP backup.

A degraded array is also an array someone has to notice. A disk failing on a Friday evening without an alert leaves a RAID 5 without redundancy all weekend. Watching disks and controllers is one of the server monitoring KPIs to track first.

Hardware RAID or ZFS

This calculator covers RAID from a hardware controller or mdadm on Linux. If ZFS manages your disks, as on TrueNAS or Proxmox VE, the rules change: RAIDZ rounds each block and ZFS keeps its own reserve. The ZFS / RAIDZ calculator accounts for this, and also estimates the space of a backup datastore.

Frequently asked questions

How much usable space does a 4-disk RAID 5 give?

Three disks out of four, or 75% of raw capacity: one whole disk goes to parity. With four 8 TB disks, 24 TB remain usable, which the system shows as 21.83 TiB. The array survives the failure of a single disk.

RAID 5 or RAID 6 with large disks?

With multi-terabyte disks, prefer RAID 6. After one failure, a RAID 5 has no redundancy left and must read every other disk without error. For four 8 TB disks, that is 24 TB to read: with the datasheet value of a consumer disk (1 error per 10¹⁴ bits), the probability of doing so without a read error drops to 14.7%. That is a pessimistic bound: the datasheet gives a maximum, and real disks do much better. So RAID 5 is not “dead”, but it has no margin left during a rebuild, whereas a RAID 6 keeps one parity.

How long does a disk rebuild take?

At least the disk size divided by the rebuild write speed. An 8 TB disk rebuilt at 100 MB/s takes 22 h 13 min, not counting production load, which often slows the rebuild down. All that time, the array runs degraded.

Why is RAID 10 faster at writing than RAID 5?

Because a random write costs two disk operations in RAID 10 (one per copy), against four in RAID 5 (read data and parity, write data and parity) and six in RAID 6. With four disks, RAID 10 therefore gives twice the write IOPS of RAID 5, at the cost of 50% efficiency instead of 75%.

Does RAID replace a backup?

No. RAID protects against a disk failure, not against a deletion, ransomware, a failing controller or losing the server: it instantly replicates the mistake to every disk. You need a separate copy, ideally offsite and impossible to modify from the protected server.

Contact

Address

5 B RUE DES NOYERS, 95300 PONTOISE, FRANCE

Let's Talk About Your Project

15 minutes to understand your needs, no commitment.

Book a Slot