ACID describes atomicity, consistency, isolation, and durability: guarantees that preserve reliable table state when multiple jobs and tools read or write shared analytical data. section 2.6.1
The guarantees address distinct failure and concurrency risks:
- Atomicity commits each write completely or not at all, preventing partial updates. section 2.6.1
- Consistency moves the dataset between valid states while preserving schema rules and table constraints. section 2.6.1
- Isolation makes concurrent operations respect ordering and conflict detection; stale writes are rejected rather than silently overwriting committed changes. section 2.6.1
- Durability keeps committed changes recorded and recoverable after failures. section 2.6.1
Concurrent lakehouse updates
Without transactional guarantees, concurrent jobs can create conflicting updates, duplicate records, or unreliable query results. Snapshot-based transactions preserve a consistent sequence of table states, allowing multiple writers and readers to work without external locking or coordination. section 1.5 section 2.6.1
In an Apache Iceberg lakehouse, ACID guarantees turn low-cost object storage into a reliable analytical system. Commit-time correctness supports shared, multi-engine access while avoiding temporary-file workarounds, external locks, and conflict-avoidance scheduling. section 1.5 section 2.6.1
