There are both known and potential opportunities to improve performance across the Java driver. This epic covers investigating those opportunities and implementing the optimizations.

Current high-impact candidates include POJO/Record codecs (a very common encode/decode path for applications) and RawBsonDocument. We currently have no POJO/Record codec benchmark coverage and therefore no time-series visibility for that path, so regressions or improvements can go unnoticed.

Key considerations:

• Benchmark coverage and time-series tracking for POJO/Record codecs and RawBsonDocument to detect regressions and validate wins.

• Benchmarks that run through common frameworks (e.g., Spring Data, Quarkus), since many applications consume the driver via frameworks and performance characteristics can differ from driver-only microbenchmarks.

• Running benchmarks across multiple Java versions, since optimization impact varies by JDK (e.g., custom little-endian encode/decode can be faster on JDK 8 but worse on JDK 11+ due to vectorization/intrinsics).

//TBD