Materialized Views in Hive for Iceberg Table Format
Cloudera
FEBRUARY 8, 2024
The snapshotId of the source tables involved in the materialized view are also maintained in the metadata. Thus, the scans and joins of the three tables in the original query are not needed and this can improve performance significantly due to both I/O cost saving and the CPU cost saving of computing the joins and aggregations.
Let's personalize your content