Configuring Query Planner

1. Setting random_page_cost

random_page_cost = 1.1     # SSD (default 4.0 for HDD)

seq_page_cost = 1.0        # baseline

cpu_tuple_cost = 0.01      # per-row processing cost

cpu_index_tuple_cost = 0.005

cpu_operator_cost = 0.0025

Parameter	Lower → Planner Prefers
random_page_cost	Index scans
seq_page_cost	Seq scans
cpu_*	Index/hash over many-row processing

effective_io_concurrency = 200     # SSD; helps Bitmap Heap Scan prefetch

SET enable_seqscan = off;     -- debugging only; never in production GUC

SET enable_indexscan = on;

SET enable_bitmapscan = on;

SET enable_hashjoin = on;

SET enable_mergejoin = on;

SET enable_nestloop = on;

Warning: enable_* are diagnostic toggles. Disabling them adds a huge cost penalty; planner may still pick a Seq Scan if it's the only option.

enable_parallel_append = on
enable_parallel_hash   = on

max_parallel_workers_per_gather = 4
max_parallel_workers            = 8
max_parallel_maintenance_workers = 4

jit                    = on
jit_above_cost         = 100000
jit_inline_above_cost  = 500000
jit_optimize_above_cost = 500000

Note: JIT helps OLAP queries with millions of rows + complex expressions. Often a net loss on OLTP.