Query and Latency Optimization

A slow query is a symptom with a location. Find the layer first, prove it with a measurement, change one thing, and measure the same workload again. Anything else is guessing with a schema change attached.

When to use

• A p95 or p99 latency regression that traces to database work.
• Detection of N+1 query patterns through statistics or SQL log analysis.
• Offset pagination that degrades at depth, or a SELECT count(*) on a hot path.
• Connection pool wait time, throughput ceilings, or lock contention symptoms.
• Choosing between a projection read path, batching, or a cache layer.
• Turning a change that felt faster into a change that is proven faster.

When not to use

• SQL dialect, index type theory, and plan node interpretation: hand off to spring-boot-postgresql.
• Entity mappings, associations, fetch annotations, and dirty checking: hand off to spring-boot-hibernate.
• Repository interfaces, Specifications, and the pagination contract: hand off to spring-boot-data-jpa.
• Metric, trace, and dashboard infrastructure: hand off to spring-boot-observability.