Skip to content

Table Format

Covers Delta Lake, Hudi, and Iceberge.

Topics

Topic What it covers
Delta Lake A transaction log layered on top of plain Parquet files in object storage — turning a folder of immutable files with no atomicity…
Hudi The table format built from the ground up around one central use case: incremental, upsert-heavy pipelines (CDC ingestion, slowly changing…
Iceberge A table format built around a different core idea than Delta Lake's linear log: a tree of metadata files (snapshots → manifest lists →…