Iceberg Datalake tables

Iceberg provides advantages over other table formats in that it manages individual data files instead of directories. This design allows data files to be written in-place and are only added to the table through an explicit commit that ensures isolated reads and writes. Data files are immutable after being written. Deleted and updated rows are written to separate data files.

Changes made to an Iceberg table result in the creation of a new metadata file (a new snapshot) referencing the current schema, partition definitions, properties, and data files. This design ensures that at any point in time, a table’s schema and exact data files are known resulting in such features as schema evolution, time travel and rollback, and data compaction to be supported. These features make a compelling case for using the table format.

Current support includes creating and dropping Iceberg tables, including create table as select (CTAS), adding externally created Iceberg tables to Db2, read, and insert support. With CTAS, you can easily create and populate an Iceberg table with data from existing Db2 or Datalake tables.