7 × 20 TB in RAID 5: the numbers
| Usable space | 120 TB (109.1 TB in Windows) |
|---|---|
| Raw space | 140 TB |
| Used for protection | 20 TB |
| Efficiency | 86% |
| Drives that can fail | Any 1 |
| Rebuild at 150 MB/s | 37 h |
| Rebuild at 50 MB/s (busy) | 4.6 days |
| Data read during rebuild | 120 TB |
| Read error risk, desktop drives | over 99% |
| Read error risk, NAS drives | 62% |
Rebuild
Replacing a failed drive means writing a full 20 TB drive. At 150 MB/s with no other load that takes about 37 h; on a busy array running at 50 MB/s, about 4.6 days. The rebuild reads every surviving drive: 120 TB in total.
Read errors
Reading 120 TB, the chance of at least one unreadable sector is over 99% with desktop drives (1 per 1014 bits) and 62% with NAS drives (1 per 1015). Nothing is left to correct it, so the affected file, or on some controllers the whole rebuild, fails. These are worst-case datasheet rates; real drives usually do better.
Our take
With 20 TB drives we would pick RAID 6. 8 drives in RAID 6 give the same 120 TB and survive two failures, or a read error during a rebuild. Keeping 7 drives, RAID 6 gives 100 TB.