RAID capacity and resilience calculator
Enter the number of disks, the disk capacity and the RAID level: usable space, efficiency, how many disk failures it survives and the probability of hitting an unrecoverable read error (URE) during a rebuild are calculated. Your data never leaves the browser.
Striping with double parity; at least 4 disks, survives two disk failures.
Data/parity disks that take part in the array.
Raw capacity from the label; pick the unit below.
Idle disks waiting in the array; not added to capacity.
The value from the vendor datasheet, e.g. 10^14 (consumer), 10^15 (enterprise SATA/NL-SAS), 10^16 (SAS). Enter the exponent E.
Sustained write speed to the new disk; leave empty to skip the time estimate.
Calculated in your browser; the values you enter are not sent.
Result
- Usable capacity
- 48 TB · 43.66 TiB
- Raw capacity (spares excluded)
- 64 TB · 58.21 TiB
- Capacity spent on redundancy
- 16 TB · 14.55 TiB
- Efficiency
- 75 %
- Efficiency (spares included)
- 66.7 %
- Disks
- 8 + 1 spare = 9 physical disks
- Tolerated disk failures
- 2 disk(s), in every case
- Minimum disks for this level
- 4
Two parity blocks are updated, so the small-write cost is higher still than RAID 5; in return it survives two failures and still has redundancy after the first one.
A hot spare starts the rebuild automatically after a failure and shortens the time without redundancy; it adds no capacity, and a second spare does not mean a second failure is tolerated.
Rebuild risk
- Data read in the no-redundancy step
- 6 disk(s) · 48 TB · 43.66 TiB
- Probability of at least one unrecoverable read error (URE)
- 31.9 %
- Rebuild time (lower bound)
- 14.81 h
RAID 6 and 60 still carry parity after the first failure; the calculation is for the data read when a second disk of the same group has also failed (the step with no redundancy). A URE during the first rebuild is normally corrected by the second parity.
New-disk capacity ÷ speed. Under load and with full disks it takes noticeably longer.
Compare levels with the same disks
| Level | Usable | Efficiency | Disk failures (min–max) |
|---|---|---|---|
| RAID 0 | 64 TB · 58.21 TiB | 100 % | 0 |
| RAID 1 | 8 TB · 7.28 TiB | 12.5 % | 7 |
| RAID 5 | 56 TB · 50.93 TiB | 87.5 % | 1 |
| RAID 6 | 48 TB · 43.66 TiB | 75 % | 2 |
| RAID 10 | 32 TB · 29.1 TiB | 50 % | 1–4 |
| RAID 50 (grouped) | 48 TB · 43.66 TiB | 75 % | 1–2 |
| RAID 60 (grouped) | 32 TB · 29.1 TiB | 50 % | 2–4 |
RAID 50/60 use the group count above.
What we calculate and how
- Usable disks: RAID 0 = n · RAID 1 = 1 · RAID 5 = n−1 · RAID 6 = n−2 · RAID 10 = n/2 · RAID 50 = n−g · RAID 60 = n−2g (n disks, g groups).
- Efficiency = usable disks ÷ n. The second figure adds the spares to the denominator.
- URE probability P = 1 − e^(−B/R). B = bits read in the no-redundancy step (disks read × capacity × 8), R = 10^E bits.
- Disks read: RAID 1/10 → 1 (the partner) · RAID 5/50 → group−1 · RAID 6/60 → group−2.
- TB = 10^12 bytes, TiB = 2^40 bytes (≈ 1.0995 × 10^12). Rebuild time = disk capacity ÷ speed.
This tool is a planning estimate. Formatting and file-system overhead, controller or software spare and stripe behaviour, correlated failures of disks from the same batch and the performance loss during a rebuild are not modelled. Vendor URE figures are often pessimistic upper bounds; real rates can differ. RAID is not a backup: deletions, ransomware and array errors are mirrored.
Shall we review your storage architecture, your redundancy level and your backup strategy together?
Request a meeting01
How to use
A
Pick the RAID level; enter the number of disks (spares excluded) and the capacity of one disk. Choose TB or TiB to match the disk label.
B
Enter the number of spare disks and the URE rate from your disk's datasheet (for example E = 15 for 1 in 10^15). You may also enter the rebuild speed.
C
Read the usable space, efficiency and tolerated disk failures; look at the rebuild risk and the level comparison table to compare the options.
02
How RAID levels change capacity and resilience
RAID 0 uses all the capacity but offers no redundancy. RAID 1 and RAID 10 mirror the data: half of the capacity (even less in an n-way RAID 1 mirror) goes to redundancy, and in return no parity calculation is needed on writes. RAID 5 gives up one disk and RAID 6 two disks of capacity to parity; efficiency grows with the disk count, but so does the amount of data read in a rebuild.
RAID 50 and RAID 60 stripe several parity groups. With many disks they limit rebuild time and risk to the group size, but every group carries its own redundancy: in the best case more disk failures are tolerated across groups, and in the worst case too many failures in a single group lose the whole array. That is why the tool shows both the minimum and the maximum tolerance.
03
Rebuilds and UREs: why we use a number
When a disk fails and a new one is inserted, the array reads the data on the remaining disks from start to end to regenerate the missing data. In this step, where redundancy has run out, a single unreadable sector (a URE) means that data is lost. With large disks the number of bits read is very high, so even a rate like 10^14 from a datasheet can point to a meaningful probability.
The tool calculates the probability as P = 1 − e^(−B/R). This model assumes errors are independent and evenly spread: in practice drives often behave better than the datasheet rate, but heat, ageing and disks from the same batch can correlate failures. Use the result as a yardstick for comparing options, not as an exact forecast; for critical data think of RAID 6/60, regular scrubbing and a separate backup together.
04
RAID, backup and hot spares
RAID gives continuity against hardware failure; it is not a backup. Accidental deletion, corrupt writes, ransomware and controller faults are mirrored or written into parity. You need a backup in a separate environment, preferably with an immutable copy.
A hot spare starts the rebuild automatically after a failure, which shortens the time without redundancy; it is not added to the capacity. In grouped levels the spare can serve all groups. We give no numbers for write cost: the real effect depends on the controller, the cache, the stripe size and the workload, and has to be measured.
FAQ
- Are the values I enter sent anywhere?
- No. The calculation runs as code in your browser; the values are not sent to any server or stored.
- Why does the capacity I calculate differ from what my operating system shows?
- Vendors count a TB as 10^12 bytes while operating systems often show TiB (2^40 bytes); the tool gives both. Formatting, the file system, metadata and space reserved by the controller are also deducted; that overhead is not included here.
- How should I choose between RAID 5 and RAID 6?
- RAID 5 survives one disk failure and leaves more capacity; RAID 6 survives two. Because rebuilds take long with big disks and many-disk arrays, RAID 6, RAID 10 or RAID 60 is usually preferred for valuable data. The URE probability in the result helps compare them.
- Why is the tolerance for RAID 10 a range?
- One disk of every mirror pair may fail, but if both disks of the same pair fail the array is lost. The guaranteed value is 1 disk; if the failures fall on different pairs it goes up to n/2 disks. RAID 50 and 60 apply the same logic to groups.
- Does a hot spare increase capacity?
- No. A spare waits idle; when it replaces a failed disk the rebuild starts automatically. That is why it is not part of usable space and only enters the denominator of the second efficiency figure.
- What URE rate should I enter?
- Use the "unrecoverable read error" value in your disk's datasheet: E = 14 for 1 in 10^14 bits, 15 for 10^15, 16 for 10^16. If you do not know it, 15 is a reasonable start; try different values of E to see how sensitive the result is.
Is your storage infrastructure resilient?
Let us review your disk arrays, your redundancy levels and your backup and restore times together. In a free discovery call we discuss where you stand and the likely risks.