1Z0-184-25 Question 65
Select 2You are designing an e-commerce platform on Oracle Cloud Infrastructure (OCI) that stores product image embeddings in an Oracle Autonomous Database with vector search capabilities. The system needs to efficiently retrieve the top-10 most similar product images for a given input embedding. Which two approaches should you implement to optimize query performance for similarity searches at scale?
- A
Use Oracle Database's approximate nearest neighbor (ANN) vector index to reduce search latency on large embedding datasets
- B
Store all product images as large BLOBs in the same table as the embeddings without any indexing
- C
Partition your embeddings table based on the embedding dimension to allow partial dimension queries
- D
Ensure embeddings are stored in a vector datatype and maintain consistent dimensionality across all records
Show answer and explanation
Correct answers: A, D
Explanation
When deploying similarity search in Oracle Autonomous Database, it's critical to use an ANN vector index (available in Oracle Database releases that support vector searches) to speed up the process of finding the nearest neighbors among millions of vectors. Storing embeddings in the native vector datatype ensures compatibility with the vector index. Refer to Oracle� documentation on 'Vector Search in Database 23c' for details on setting up the index and best practices for embedding storage.
- A. Correct.
Correct. Oracle Database supports an ANN vector index that accelerates similarity searches by quickly narrowing down the set of potential matches, significantly improving performance over purely brute-force methods. Enabling this specialized index is a best practice for large-scale vector queries.
- B. Incorrect.
Incorrect. Storing all product images as BLOBs in the same table as embeddings without indexing is inefficient. Because these BLOBs do not contribute to the vector-based similarity calculation, mixing them with embedding data in a single table without indexing would hurt query performance and complicate data retrieval.
- C. Incorrect.
Incorrect. Partitioning by embedding dimension is not a supported or beneficial strategy. Partitioning typically involves data ranges or lists, not individual dimensions of a vector. Such an approach would introduce unnecessary complexity without improving query performance in vector search scenarios.
- D. Correct.
Correct. Consistent dimensionality is critical for valid vector comparisons, and using the vector datatype allows the database to perform efficient similarity calculations. Any mismatch in dimensions or data types can lead to errors or degrade search performance.