Custom JDBC will be the best in this case, here is why:

• Custom JDBC executes SQL directly, reducing latency and memory usage.

• JPA (Hibernate) introduces ORM overhead, persistence context tracking, and automatic flushing, which significantly slow down bulk operations.

• JDBC Batch Processing allows optimized bulk inserts/updates, while JPA requires additional configurations.

References:

• Stack Overflow discussion reports JDBC performing 10-20x faster than Hibernate for bulk inserts.

Bulk insert: JdbcTemplate vs JPA

• Oracle Forums test case found JDBC inserted 15,000 records in 39 seconds, while JPA took 2 minutes 30 seconds due to ORM overhead.

https://forums.oracle.com/ords/apexds/post/is-jpa-effecient-enough-to-use-bulk-insert-using-jpa-0711

• Medium article highlights how Hibernate’s entity tracking and auto-flushing make it unsuitable for high-volume transactions.

https://medium.com/%40wahyaumau/boost-jpa-bulk-insert-performance-by-90-3a6232d9068d