RAID Rebuild Time Calculator

When a disk fails, the array rebuilds its contents onto a replacement. Time is roughly disk size divided by rebuild speed:

hours = disk size (bytes) / rebuild speed (bytes per second) / 3,600

Production load slows it down: at 30% load, the rebuild gets about 70% of the speed, so the time is divided by 0.7.

URE exposure is the chance of hitting at least one unrecoverable read error while reading one full disk. Drive makers quote a URE rate as 1 error per 10¹⁴ or 10¹⁵ bits read. The calculator treats errors as random and uses 1 - e^{-bits read / 10^exponent}.

The default is a 12 TB disk rebuilding at 180 MB/s under 30% production load, on drives rated 1 URE per 10¹⁴ bits:

• Time: 12,000,000 MB / 180 MB/s = 66,667 s = 18.52 hours at full speed
• Under 30% load: 18.52 / 0.7 = 26.5 hours (1.10 days)
• Bits read: 12 x 10¹² x 8 = 9.6 x 10¹³
• URE exposure: 1 - e^-0.96 = 61.71%

The same disk rated at 10¹⁵ gives 9.15%. That gap is a big reason RAID 6 is preferred over RAID 5 for large disks.

Start with six equal 12 TB disks in RAID 6. The capacity calculator shows 72 TB raw and 48 TB usable, with two-drive failure tolerance under its stated model. Hot spares, filesystem and controller overhead are not included.

Now open RAID Rebuild for one 12 TB replacement disk. At 180 MB/s sustained rebuild speed and 30% production load, the model gives 26.5 hours. With a URE exponent of 14, it reports 61.71% per-disk unreadable-sector exposure. That figure is not the chance the array fails and does not model the controller's recovery behavior.

Use the rebuild estimate to ask whether the degraded window is acceptable. Then plan an independent backup and test restoration. RAID tolerance, a rebuild estimate and a backup-capacity estimate answer different questions.

No. It covers reading one disk. A rebuild reads every surviving disk, and RAID 6 can often correct a read error that would stop a RAID 5 rebuild. Treat it as a way to compare drive ratings, not a failure forecast.

They are spec-sheet limits, and many drives do better in practice. They are still useful for comparing drive classes: desktop drives are usually rated 10^14 and enterprise drives 10^15 or better.

Controllers often throttle rebuilds to protect production traffic, and SMR drives or heavy random I/O can cut speed a lot. Check the rebuild rate setting on your controller or NAS.