1Z0-184-25 Question 30
Single answerYou are running an Oracle Database 23c-based service in OCI that supports vector columns. Your table IMAGE_EMBEDDINGS stores image embeddings in a VECTOR column EMBEDDING_COL of dimension 256, which you use for similarity searches. The data scientists have upgraded the embeddings to dimension 512 and require minimal downtime. How should you update the table schema to accommodate the new dimension and preserve existing data?
- A
Use an offline table re-creation statement to drop and recreate the entire table with the new dimension.
- B
Issue a direct ALTER TABLE ... MODIFY statement to change the dimension of EMBEDDING_COL from 256 to 512 in place.
- C
Use an approach where you rename the existing column, add a new VECTOR column with dimension 512, migrate data from the old column to the new one, and rebuild the vector index.
- D
Create a materialized view to hold the new vector embeddings, then refresh it back into the original table via a complete refresh.
Show answer and explanation
Correct answer: C
Explanation
When modifying the dimension of a VECTOR column in Oracle Database 23c, there is no direct DDL command to change its dimension in place. According to best practices and official documentation on vector data types, you must create a new column with the updated dimension, migrate your data, and index the new column for similarity searches. Renaming the original column ensures minimal disruption to production activities. After verifying data integrity, you can remove the old column and continue normal operations without significant downtime.
- A. Incorrect.
Incorrect. Dropping and recreating the entire table would cause significant downtime and potential data loss, making it impractical for a production environment.
- B. Incorrect.
Incorrect. There is no direct option to modify the VECTOR dimension in place. Attempting to do this would result in an error. Dimension changes require recreating the column.
- C. Correct.
Correct. Renaming the old column and then adding a new one with dimension 512 allows you to migrate and transform data without impacting ongoing operations. Rebuilding the index ensures updated similarity search capabilities.
- D. Incorrect.
Incorrect. A materialized view does not solve the underlying need to change the dimension in the base table. While it can replicate data, it will not allow direct schema changes in the original table.