Transactional mode

Each record that is ingested into Data Lake is converted as an independent row. No evaluation is performed to determine whether new records update existing converted records. As a result, duplicate or multiple versions of the same logical record in Data Lake are returned directly in query results.

Because no record comparison and deduplication are performed during conversion, conversion performance remains consistent for frequently ingested data and for records that are updated multiple times between query runs. Therefore, transactional mode is appropriate for scenarios with frequent query runs or highly dynamic update patterns.

Any required deduplication or record consolidation must be performed outside the conversion process with one of these methods:

• Through query logic.
• Through deduplication in downstream systems in which query results are loaded and processed.