What this calculator works out
This calculator states how many drive failures each RAID level survives for a given drive count, with a note where the answer depends on which drives fail.
The number is a floor, not a guarantee. RAID 10 can sometimes survive several failures and sometimes only one, depending entirely on where they occur.
Enter your details
Before you rely on the figure
- RAID 0 survives no failures. It is included because it is offered and frequently chosen by accident.
- RAID 5 survives one; RAID 6 survives two, regardless of drive count.
- RAID 1 survives until one copy remains, so a three-way mirror survives two failures.
- RAID 10 survives at least one, and more only if the failures fall in different mirror pairs.
An array running with a failed drive is not merely degraded — it is in its most vulnerable state, and rebuilding onto a replacement stresses every remaining drive for hours or days. Drives bought together and used identically tend to fail at similar times, which is precisely why a second failure during a rebuild is more likely than the raw statistics suggest.
How this calculator works
Tolerance by level:
RAID 0: 0 drives
RAID 1: drives − 1
RAID 5: 1 drive
RAID 6: 2 drives
RAID 10: at least 1, more only across different mirror pairsRAID 5 and 6 tolerate a fixed number regardless of array size, because parity protects the stripe rather than the drive. Adding more drives to a RAID 5 array increases capacity and increases the chance that one of them fails, without increasing what the array can survive.
Worked example: four drives at RAID 0
Using the default figures — RAID 0 across 4 drives:
- Drives that can fail: 0
Zero, and worse than a single drive — four drives striped means four independent chances of total loss rather than one. The same four drives at RAID 6 survive any two failures. If a system was set up by someone else and the level is unknown, this is the first thing worth checking, because RAID 0 is frequently the default on consumer enclosures and is chosen by people who believe RAID means protection.
Common mistakes
- Assuming any RAID level offers protection. RAID 0 offers none.
- Adding drives to RAID 5 for safety. It raises capacity and risk together, not tolerance.
- Reading RAID 10's figure as guaranteed. It depends on which drives fail.
- Delaying a rebuild. A degraded array has no remaining margin.
- Buying all drives from one batch. They tend to fail at similar times.
Frequently asked questions
What happens during a rebuild?
The array reconstructs the failed drive's data onto a replacement by reading everything on the remaining drives. On large drives this takes many hours or several days, during which the array has no redundancy — or reduced redundancy on RAID 6 — and every remaining drive is under sustained load. This is the single most dangerous period in an array's life.
Should I use a hot spare?
It shortens the time between failure and rebuild start, which is worthwhile where nobody is watching the system continuously. It does not shorten the rebuild itself. For a small array, RAID 6 with no spare is frequently better than RAID 5 with one, because it survives a failure during the rebuild rather than merely starting it sooner.
How do I know a drive has failed?
The controller or NAS should alert, and configuring those alerts is essential — an unnoticed failed drive means an array running without redundancy indefinitely. SMART monitoring gives advance warning in some cases. Scheduled scrubs, which read the whole array to detect errors before they matter, are worth enabling.
Does RAID replace a backup?
No. Three copies of the data, on at least two types of media, with one kept off-site. Each part defeats a different failure: drive failure, accidental deletion, and fire, theft or ransomware.
Is what I enter stored?
No. Figures are processed in your browser and never transmitted or retained.
Related tools
References
- National Cyber Security Centre — resilience, backup and recovery guidance
- Information Commissioner's Office — availability and integrity obligations for personal data
Sources are checked at publication and can change — how I choose and check references.