1Z0-184-25 Question 26
Single answerYour data science team has retrained a recommendation model and produced new vector embeddings. These embeddings must replace existing data in the 'ITEM_VECTORS' table within your Oracle Autonomous Database instance on OCI. The table has a vector index configured to support similarity search. You want to minimize downtime and ensure no loss of query performance during the update. Which approach is recommended to keep the vector index in sync while performing the DML operations?
- A
Directly run an UPDATE on the vector column for all rows in a single large transaction. The index automatically stays up to date without additional actions.
- B
Use incremental batched UPDATE statements on the vector column, then issue an ALTER INDEX ... REBUILD or ALTER INDEX ... REBUILD ONLINE once all updates complete.
- C
Drop the existing vector index, update the vectors using INSERT INTO ... SELECT, and then recreate the vector index after the data is updated.
- D
Perform a MERGE statement combining both old and new embeddings row by row, which automatically refreshes the vector index without extra steps.
Show answer and explanation
Correct answer: B
Explanation
When performing large-scale DML updates on columns that contain vector data, best practice is to batch the updates to manage concurrency and then explicitly rebuild or synchronize the vector index. Refer to Oracle� documentation for vector search in Autonomous Database for guidance on using 'ALTER INDEX ... REBUILD' or 'ALTER INDEX ... REBUILD ONLINE' commands to maintain performance after significant changes to embedding values.
- A. Incorrect.
Option 1 is incorrect. While Oracle Database 23c and later can handle some incremental index maintenance for vectors, large bulk updates may require explicit index synchronization or rebuilding to maintain optimal performance. Relying solely on an automatic process without additional index maintenance can cause degraded query performance.
- B. Correct.
Option 2 is correct. Batching the UPDATE statements helps manage concurrency and minimize locking overhead. After all updates finish, running an index rebuild or online rebuild ensures the vector index is fully synchronized, preserving fast similarity queries.
- C. Incorrect.
Option 3 is incorrect. Although dropping the index and recreating it afterward ensures accurate indexing, it causes longer downtime and potential performance issues during the reindexing process, which is not optimal for minimizing service interruptions.
- D. Incorrect.
Option 4 is incorrect. A MERGE by itself does not fully refresh the vector index automatically. Oracle� vector index maintenance generally requires an explicit rebuild step unless you employ specialized online synchronization methods, which this option does not address.