The RAID calculator above takes a level, a drive count and a drive size, and returns usable capacity, storage efficiency, how many drives can fail without losing data, the write penalty the level imposes, and an estimate of how long a rebuild will take. Every figure is computed in your browser; nothing is sent anywhere.
Arb Digital publishes this alongside its other free developer and infrastructure tools because array design is a set of trade-offs that are easy to state and easy to get wrong. Capacity, redundancy and write performance pull against each other, and the level that looks most efficient on a spreadsheet is often the one that will not survive the rebuild after its first failure.
What This RAID Calculator Does
RAID combines several drives into one logical volume, distributing data across them for capacity, performance, redundancy or some mixture of the three. Each level makes a different bargain. Striping alone gives you every byte you paid for and no protection. Mirroring gives you half the capacity and immediate protection. Parity levels sit between the two, spending the equivalent of one or two drives on recovery information.
The calculator applies the standard capacity formula for the level you choose, reports the efficiency as a percentage, states the fault tolerance in drives, and shows the write penalty — the number of physical input/output operations required to complete one logical write. That last figure is the one most often left out of planning, and it is why a parity array can feel slow on a workload full of small random writes even when its sequential throughput looks excellent.
It also estimates rebuild time from drive size and throughput, because rebuild duration is the window during which a degraded array is exposed to a second failure. That risk is the deciding factor between RAID 5 and RAID 6 on large modern drives.
How to Use It
- Pick a level. The minimum drive count and the tolerance update with it, and the tool will say if the count is too low.
- Enter the drive count and size. Use the size printed on the drive; the conversion to what the operating system reports is shown below the result.
- Set the parity groups for RAID 50 or 60. More groups means more parity overhead and better tolerance.
- Add hot spares if you plan to fit them. They shorten the time before a rebuild starts but do not add usable capacity.
- Check the rebuild estimate. If it runs into days, that is a strong argument for dual parity or mirroring rather than single parity.
The Formulas for Each Level
Writing n for the number of drives and s for the size of the smallest drive:
RAID 0: usable = n × s, tolerance 0, write penalty 1.
RAID 1: usable = s, tolerance n − 1, write penalty 2.
RAID 5: usable = (n − 1) × s, tolerance 1, write penalty 4.
RAID 6: usable = (n − 2) × s, tolerance 2, write penalty 6.
RAID 10: usable = (n ÷ 2) × s, tolerance at least 1 and up to n ÷ 2 if the failures land in different mirrors, write penalty 2.
RAID 50: usable = (n − g) × s for g parity groups, tolerance one drive per group.
RAID 60: usable = (n − 2g) × s, tolerance two drives per group.
Work through the default: six drives of 4 TB in RAID 5. Raw capacity is 24 TB. Usable is (6 − 1) × 4 = 20 TB, so efficiency is 83.3 per cent, one drive can fail, and every small write costs four operations — read the old data, read the old parity, write the new data, write the new parity. Switch to RAID 6 and usable drops to 16 TB at 66.7 per cent efficiency, two drives can fail, and the write penalty rises to six. Switch to RAID 10 and usable is 12 TB at 50 per cent, with a write penalty of two and much better small-write performance.
Rebuild Risk: Why RAID 5 Fell Out of Favour
When a drive in a parity array fails, the array is degraded but still serving data. Recovery means reading every block on every surviving drive and reconstructing the missing one. Two things make that dangerous on large drives.
The first is duration. Rebuilding a large drive means writing its entire capacity while the array continues to serve production traffic, and that can take many hours or several days. Throughout that window a RAID 5 array has no redundancy at all: a second failure loses the volume. The tool's rebuild estimate is deliberately prominent for this reason.
The second is read errors. Reconstruction requires every remaining block to be readable. Drive manufacturers publish an unrecoverable read error rate on their data sheets, and the total number of bits that must be read during a rebuild grows with both drive size and array size. On a large single-parity array, the chance of meeting an unreadable sector part-way through a rebuild is no longer negligible, and on many controllers that ends the rebuild.
Dual parity exists precisely to cover that window: RAID 6 survives a second failure and can still reconstruct around a bad sector while degraded. Microsoft's documentation on fault tolerance and storage efficiency in Storage Spaces works through the same trade-off in a software-defined form, noting that dual parity provides the same fault tolerance as three-way mirroring with better storage efficiency, and explicitly discouraging single parity because it tolerates only one failure at a time.
The Write Penalty Nobody Budgets For
A logical write to a parity array is not one physical operation. For RAID 5, changing part of a stripe requires reading the old data, reading the old parity, computing the new parity, then writing both — four operations for one write. RAID 6 maintains two independent parity values, so the same update costs six.
The practical consequence is that the input/output budget of a parity array is much smaller than the sum of its drives suggests for random-write workloads. If a database needs a certain number of write operations per second, divide the array's raw capability by the penalty before deciding whether the design is adequate. Mirroring costs only two operations per write, which is why transactional databases and virtual machine hosts are so often built on RAID 10 despite its fifty per cent capacity cost.
Sequential writes behave much better, because a full-stripe write computes parity from data already in hand and skips the read phase entirely. That is why the same array can look fast when copying large files and slow under a busy database, and why a benchmark that only tests sequential throughput tells you almost nothing about how the array will behave in production. For sizing the network side of the same system, the bandwidth converter and data storage converter handle the unit arithmetic.
Why RAID Is Not a Backup
This deserves stating plainly, because it is the most expensive misunderstanding in the field. RAID protects against exactly one failure mode: a drive dying. It replicates every write faithfully, including the wrong ones. If a file is deleted, it is deleted on every member instantly. If ransomware encrypts a volume, the encryption is mirrored. If a controller or a memory fault writes corrupt data, the corruption is written with full redundancy. Fire, flood, theft and a datacentre outage take every member together, because they are in the same chassis.
A backup is a separate copy, on separate media, that can be restored to a point in time before the damage. The widely used guideline is three copies of the data on two different media with one copy off-site. NIST's Special Publication 800-34 Revision 1, Contingency Planning Guide for Federal Information Systems, sets out how to plan for information system recovery as a process with tested procedures rather than a hardware feature. RAID buys uptime through a drive failure. It does not buy recoverability, and treating it as though it does is how organisations discover the difference at the worst possible moment. If uptime is what you are measuring, the uptime SLA calculator converts availability targets into permitted downtime, and the CRC32 hash generator is useful for verifying that copies match after a restore.
Arb Digital builds and hosts fast websites with real off-site backups and tested restores, not just redundant drives.
See Web Design Services Talk to Arb DigitalCommon Mistakes to Avoid
- Treating RAID as a backup. It replicates deletion, corruption and ransomware perfectly. Keep separate copies, including one off-site, and test restoring them.
- Choosing single parity for a large array. The rebuild window on big drives is long, and a second failure or one unreadable sector during it can end the volume.
- Ignoring the write penalty. A parity array delivers far fewer random writes per second than its drive count suggests. Divide by four for RAID 5 and six for RAID 6 before sizing.
- Mixing drive sizes and expecting the total. Every drive contributes only as much as the smallest member, so the extra capacity on the larger ones is simply unused.
- Forgetting the decimal-to-binary conversion. A 4 TB drive shows as about 3.64 TiB. Nothing is missing; the units differ.
Related Free Tools From Arb Digital
For unit conversions between bytes, gigabytes and tebibytes, use the data storage converter, and for transfer rates the bandwidth converter. The uptime SLA calculator turns availability percentages into allowed downtime, the percentage calculator covers the efficiency arithmetic, and the CRC32 hash generator helps verify copies. If you are budgeting the hardware, the budget calculator is a useful companion. Everything else is in the free online tools hub.
Frequently Asked Questions
No. RAID protects against a drive failing and nothing else. Deletions, ransomware, corruption, theft and fire all affect every member of the array at once. A backup is a separate copy on separate media that can be restored to an earlier point in time.
The capacity of all drives minus one, so six 4 TB drives give 20 TB usable out of 24 TB raw, an efficiency of 83.3 per cent. One drive can fail without data loss.
RAID 6 costs one more drive of capacity and survives two failures, including a failure during a rebuild. On large drives, where rebuilds take many hours or days, that second parity is the difference between a degraded array and a lost one.
The number of physical operations needed for one logical write. RAID 5 costs four because it must read the old data and parity before writing both back, RAID 6 costs six, and mirroring costs two.
At least one, and up to half the array if each failure lands in a different mirrored pair. Two failures within the same pair lose the volume, so the guaranteed tolerance is one.
Drives are sold in decimal units where a terabyte is a trillion bytes, while most operating systems report binary units where a tebibyte is 2 to the power 40 bytes. A 4 TB drive therefore shows as about 3.64 TiB.
Every drive contributes only as much capacity as the smallest member, so the surplus on the larger drives is unused. Some software solutions handle mixed sizes better, but classic RAID levels do not.
No. A hot spare sits idle until a member fails, then the controller rebuilds onto it automatically. It shortens the time spent degraded, which is valuable, but it contributes nothing to usable space.
