#Postgres DBAs: cold page reads eating up query execution time on full table scans?
Some config params that help reduce I/O latency by 40%:
effective_io_concurrency = 200 (SSDs)
random_page_cost = 1.1
max_parallel_workers = 8
These changes (and others!) resulted in a 30% improvement in checksum time. Explore our latest blog for before/after EXPLAIN ANALYZE output + explanations behind each param at the DB and OS level.
https://www.pgedge.com/blog/optimising-cold-page-reads-in-postgresql