1Z0-184-25 Question 1
Single answerYou are designing a new recommendation engine on Oracle Cloud Infrastructure that uses vector embeddings for personalizing product suggestions. The engine must deliver near real-time similarity queries over millions of embedded vectors stored in Oracle Database 23c or higher. Which approach should you recommend to meet these performance and scalability requirements?
- A
Convert the vectors to numeric IDs and store them in a standard NUMBER column
- B
Use the new VECTOR data type in Oracle Database 23c with approximate nearest neighbor (ANN) indexing
- C
Store embeddings as JSON documents in Oracle Autonomous JSON Database with B-tree indexes
- D
Keep embeddings in Oracle Object Storage and retrieve them during each query
Show answer and explanation
Correct answer: B
Explanation
Oracle Database 23c provides a dedicated VECTOR data type and ANN indexing to handle high-dimensional vector queries efficiently. By using this specialized data type, you can achieve near real-time performance for similarity searches across large vector datasets. For more information, see the official Oracle Database 23c documentation on vector data processing and approximate nearest neighbor indexes.
- A. Incorrect.
Storing vectors as numeric IDs in a regular NUMBER column creates a limited representation. You lose the granularity and dimensional nature of the original vector, making similarity queries highly inefficient or impossible without large-scale conversions.
- B. Correct.
This option is correct. Oracle Database 23c introduces a dedicated VECTOR data type and an approximate nearest neighbor indexing mechanism optimized for similarity searches. This preserves the full vector structure and delivers efficient queries across large vector datasets.
- C. Incorrect.
While JSON storage is suitable for many JSON-based use cases, B-tree indexes are not designed for vector similarity queries. You would need specialized indexing to address high-dimensional data effectively, making a B-tree index suboptimal for vector embeddings.
- D. Incorrect.
Storing embeddings in Oracle Object Storage might be convenient for large file objects, but retrieving them for each query would introduce significant latency. There is no direct built-in vector index, and each query would require reading objects into memory, making real-time similarity searches impractical.