Data protection & recovery
Bit rot
Bit rot is the gradual, silent corruption of data at rest: bits on a storage medium change value over time without any write being issued and without the device reporting an error. The damaged data is returned as if it were valid unless the system has an independent way to check it.
What causes bit rot?
Storage media are physical and imperfect. Several mechanisms can alter stored data.
| Cause | How it corrupts data |
|---|---|
| Magnetic decay on hard drives | Weakening of the recorded signal over years |
| Charge leakage in flash cells | Stored charge drifts, especially at high wear or temperature |
| Latent sector errors | A region becomes unreadable and is only discovered when read |
| Misdirected or torn writes | Data lands in the wrong place or is only partly written |
| Firmware and controller faults | Wrong data returned without an error flag |
| Memory errors in transit | A bit flips in RAM before the data reaches the disk |
Why is it called silent?
Drives have internal error correction and report failures they detect. The dangerous cases are the ones that slip past: the drive returns data it believes is good, the file system passes it up and the application uses it. Nothing logs an error. The problem surfaces much later, when a file will not open, a backup will not restore or an analysis gives wrong results, by which time the corruption may have been copied into every replica and backup made since.
How is bit rot detected?
Detection needs a reference that is stored separately from the data. A checksum computed when the data was written is compared with one computed when it is read. A mismatch proves the data has changed. Because much stored data is rarely read, systems also run scrubbing: a background process that reads everything on a schedule and verifies it, so that damage is found while redundant copies are still intact. The rate matters at scale. With enough drives and enough years, undetected errors are a statistical certainty.
How is it repaired?
Detection alone identifies the damage. Repair needs redundancy. When a checksum fails, the system fetches another replica or reconstructs the piece from the remaining fragments under erasure coding, verifies the result and rewrites the bad copy. This is why checksums and redundancy are designed together, and why data durability figures assume regular scrubbing. Long-term archives, where data sits untouched for years, depend on it most. Silent data corruption and integrity verification cover the practice.
Frequently asked questions
How common is bit rot?
Rare per drive, and expected across large fleets over years. At petabyte scale it is planned for, not treated as an exception.
Does RAID protect against bit rot?
Not by default. Standard RAID recovers from a drive that fails outright. It often cannot tell which copy is wrong when a drive returns bad data without an error.
Do backups protect against bit rot?
Only if the corruption is noticed before the good backups expire. Corrupted data is backed up as faithfully as good data.
Related terms
- Checksum: the mechanism that detects changed data.
- Data durability: the property bit rot threatens.
- Erasure coding: redundancy used to repair damaged pieces.
- Data rebuild: recreating data after a detected failure.
- Annualized failure rate (AFR): the rate of outright drive failures.