P[drive failure]: how reliable is your NAS?
3 days ago
- Hardware failures are inevitable; redundancy is critical for long-term data preservation.
- The author plans a four-drive Btrfs RAID1 setup with offsite synchronization at a relative's home.
- Hard drive failures follow a bathtub curve but modern drives show relatively constant annual failure rates (AFR).
- Sector errors (LSEs) are another failure mode, with probability modeled from empirical data.
- Mathematical modeling estimates data loss probability using Poisson processes and sector error distributions.
- With four drives in RAID1, the probability of data loss over 50 years is about 1 in 10 million.
- Regular scrubbing helps but primary protection comes from redundancy: using multiple drives.
- The analysis uses qalc for unit-aware calculations and probability functions.
- A key finding: adding one redundant drive reduces failure probability by a factor of ~300.