Columnar Data, from the inside out Iceberg, Explained →

Columnar / Storage and tables

Iceberg

Iceberg is an open table format: metadata that turns a collection of Parquet files into a table with versions. Parquet describes the contents of one file. Iceberg decides which files belong to a snapshot, including which replaced files an older reader still needs. A catalog points through metadata JSON, a manifest list and manifests to the data. Engines such as Spark, Trino and Flink read and write tables through those references.

ParquetColumns stored as bytesIceberg · Delta LakeFiles selected by a versionArrowTyped arrays in memoryDuckDB · DataFusionOperators turn values into answers
Table metadata selects files. Readers decode stored columns into arrays. Engines compute on those arrays.
Read Iceberg, Explained →Inspect your table metadata →