Ryan Marcus, the developer of the experimental optimizer Bao for PostgreSQL, which utilizes machine learning to optimize query execution, has recently published the performance testing results of the full-time optimizer for PostgreSQL requests. The testing spanned across PostgreSQL branches from version 8.4 (2009) to version 16 (2023). Productivity was evaluated using the Join Order Benchmark, which consists of over 100 complex queries with numerous Join operations to assess various aspects of the query optimizer.
Comparing the performance of test queries in PostgreSQL 16 to the older version 8.4, there was nearly a twofold increase in speed. On average, each significant new version of PostgreSQL showed a 15% improvement in performance when conducting Job tests.