Skip to content

Crash Consistency

Systems must maintain valid, recoverable state even when failures occur mid-operation, requiring careful ordering of writes and recovery protocols.

Tension Addressed

Durability vs. Performance

Recognizable As

Atomic operations, journaling, two-phase commits, checksums for corruption detection, recovery modes, "torn write" discussions

Canonical Sources

Mohan et al., "ARIES" (1992); Pillai et al., "All File Systems Are Not Created Equal" (2014); Gao et al., "Crash Consistency" survey papers

Transfers To

Database transaction management, distributed consensus, file system design, any system requiring atomicity guarantees

Crashes can occur at any point, potentially leaving data structures in inconsistent states. This concept explains why "atomic" operations are valuable and complex.