PostgreSQL Configuration
shared_buffers
This parameter has the most variance of all. Some workloads work best with minimal values (such as 1GB or 2GB), even with huge database volumes. Other workloads require large values. The LEAST(RAM/2, 10GB) is a reasonable starting point.
This formula has no specific reason beyond the PostgreSQL community’s years of collective wisdom and experience. Complex interactions between the kernel cache and shared_buffers make it nearly impossible to describe exactly why this formula generally provides good results.
References
- https://info.enterprisedb.com/rs/069-ALB-339/images/Q4%202021%20-%20Webinar%20-%20Slides%20-%20Tuning%20Tips%20to%20Maximize%20Postgres%20Performance.pdf
- https://www.enterprisedb.com/postgres-tutorials/introduction-postgresql-performance-tuning-and-optimization (PostgreSQL v13)
- https://www.enterprisedb.com/blog/improving-postgresql-performance-without-making-changes-postgresql (EDB Advanced Server v13)
- https://cdrdv2-public.intel.com/816245/postgresql-on-azure-dsv5-esv5-optimization.pdf (PostgreSQL v13)
- https://cdrdv2-public.intel.com/684953/Open-Source-Database-Tuning-Guide-on-3rd-Generation-Intel-Xeon-Scalable-Processors-1.pdf (PostgreSQL v13)