Find expensive SQL
Start with statements that consume high total time, have high average latency, or sit on critical paths such as checkout, booking, reporting, and search.
SQL performance tuning
SQL performance tuning starts with workload evidence, not guesswork.
Good tuning identifies the SQL statements that affect real users, explains why they are expensive, and turns that evidence into safer changes to query shape, indexes, data access, or capacity.
Read execution evidence
Use plans, row counts, scan type, join method, sort work, and temporary writes to understand where the database spends time.
Reduce work
Filter earlier, return fewer columns, avoid repeated lookups, simplify joins, and remove unnecessary sorting before scaling infrastructure.
Review indexes
Index candidates should match predicates, joins, ordering, and selectivity. A useful index lowers read cost without creating avoidable write overhead.
Validate changes
Compare before and after plans, staging behavior, lock risk, and production rollout impact before treating a tuning idea as complete.
Keep context
Tuning one query can move pressure elsewhere. Connect SQL work to broader database optimization evidence.
Tuning workflow
Move from symptom to measured change.
Rank SQL by impact
Sort by workload cost, not the single slowest statement. Frequent medium-cost SQL can matter more than rare outliers.
Choose the smallest useful fix
Pick the change that directly addresses the observed cost: index, rewrite, pagination, batching, statistics, or capacity review.
Need engine-specific evidence later?
This page covers broad SQL performance tuning. For PostgreSQL databases, continue to the PostgreSQL optimization path or the focused PostgreSQL slow query audit after the general tuning scope is clear.