Hong Life CalculatorsUnderstanding the RAID rebuild time estimator — concept and principle
When a disk in a RAID array fails and is replaced, the controller must reconstruct the missing data onto the new disk. Until that rebuild finishes, the array runs degraded with less or no redundancy. This estimator shows how long that vulnerable window lasts, which helps when choosing a RAID level, sizing hot spares or planning maintenance.
How it is calculated
Rebuild time (hours) = disk size (bytes) ÷ rebuild rate (bytes/s) ÷ 3,600.
Disk size is in decimal terabytes (1 TB = 10¹² bytes, as drive makers label them) and the rate is in decimal megabytes per second (1 MB/s = 10⁶ bytes/s). The rate should be the sustained speed the rebuild actually achieves, not the drive's peak sequential figure.
The principle
A classic hardware or software RAID rebuild works block by block across the whole replacement disk, whether the blocks hold data or not. For each stripe it reads the surviving disks, recomputes the missing piece (a copy in RAID 1/10, parity in RAID 5/6) and writes it to the new disk. The whole capacity therefore has to pass through the slowest part of that chain, which is usually the write speed of the new disk or a rebuild-priority limit set to protect normal workloads. Dividing capacity by that sustained rate gives the minimum time. Some file-system-aware systems, such as ZFS resilvering, copy only allocated blocks, so a half-empty pool can finish sooner.
Worked example
- Inputs: 8 TB disk, 150 MB/s rebuild rate.
- Capacity in bytes: 8 × 10¹² = 8,000,000,000,000.
- Rate in bytes/s: 150 × 10⁶ = 150,000,000.
- Seconds: 8 × 10¹² ÷ 1.5 × 10⁸ = 53,333 s.
- Hours: 53,333 ÷ 3,600 = 14.81 hours.
| Rebuild rate (MB/s) | Rebuild time (hours) |
|---|---|
| 37.5 | 59.25925926 |
| 75 | 29.62962963 |
| 112 | 19.75308642 |
| 150 | 14.81481481 |
| 188 | 11.85185185 |
| 225 | 9.87654321 |
| 300 | 7.407407407 |
How to read the result
Treat the result as a best case. Arrays serving users often rebuild at a fraction of the drive's raw speed, so real times can be two or more times longer. During the whole window a RAID 5 array has no remaining redundancy: one more failure, or an unreadable sector on a surviving disk, can cause data loss. That is why large disks push designers towards RAID 6, RAID 10 or erasure coding, and why backups are still needed.
Common mistakes
- Using the drive's peak sequential speed instead of a realistic sustained rebuild speed.
- Mixing TB and TiB, or MB/s with Mb/s (megabits).
- Assuming the array is safe while rebuilding.
Frequently asked questions
Why is RAID 5 risky with large disks?
Rebuilds take longer and must read every sector of all surviving disks. Datasheets quote unrecoverable read error rates such as one per 10¹⁴ or 10¹⁵ bits, so reading many terabytes raises the chance of hitting an error.
Does a hot spare shorten the rebuild?
It does not make the copy faster, but it starts the rebuild immediately instead of waiting for someone to swap the disk.
Sources: D. A. Patterson, G. Gibson and R. H. Katz, A Case for Redundant Arrays of Inexpensive Disks (RAID), ACM SIGMOD 1988 · IEC 80000-13, Quantities and units — Information science and technology (decimal and binary prefixes)