Learn how to improve Presto query performance through runtime ANALYZE and Quick Stats in this 11-minute technical talk. Explore how to overcome the optimizer's limitations when dealing with recently ingested data by leveraging file-level metadata from Delta and Hudi formats, along with statistics from Parquet and ORC footers. Discover techniques for building representative partition-level statistics that can be cached for future queries and persisted to the metastore. Understand how sampling strategies can be implemented to enhance statistical precision, ultimately leading to better query execution plans and improved database performance.
Overview
Syllabus
Quick Stats - Runtime ANALYZE for Better Query Plans - Anant Aneja, Ahana
Taught by
Presto Foundation