What is the difference between Hibernate's first-level and second-level cache?

I add second-level caching only after measuring a repeatable database bottleneck and defining tolerated staleness. Cache hit rate alone is not success if invalidation traffic, memory pressure, or inconsistent reads rise. The cache strategy must match the entity's update pattern, and operational controls need eviction and metrics. Often a purpose-built DTO cache with an explicit key and lifetime is easier to reason about than transparently caching a mutable entity graph.