1Z0-184-25 Question 5
Single answerYou are designing a semantic search platform on Oracle Cloud Infrastructure to recommend relevant articles to users based on their reading history. Your plan is to generate vector embeddings for each article and store them for fast similarity searches. Which approach best ensures you can efficiently insert new vectors, update existing ones, and perform near real-time vector similarity lookups as your content library grows?
- A
A) Maintain a traditional B-Tree index in Oracle Autonomous Database for your vector columns
- B
B) Store vectors in external POSIX-compliant Object Storage and handle indexing in an external application
- C
C) Use Oracle Database with a specialized vector data type and approximate nearest neighbor (ANN) indexing
- D
D) Store vectors in a caching layer and perform brute-force searches in your application code
Show answer and explanation
Correct answer: C
Explanation
Using Oracle Database with a dedicated vector data type and approximate nearest neighbor indexing is the recommended approach for high-dimensional similarity searches at scale. By leveraging these native features, you can efficiently handle inserts, updates, and queries without manually maintaining indexes in external storage or adding excessive overhead in your application code. Refer to Oracle� documentation on Vector Search and indexing best practices for details on implementation and configuration.
- A. Incorrect.
A) Incorrect. B-Tree indexes are optimized for exact matches or range queries, not for high-dimensional vector similarity searches. This approach will lead to inefficient lookups and slower performance as the number of embeddings grows.
- B. Incorrect.
B) Incorrect. While Object Storage is highly scalable for unstructured data, it lacks built-in vector indexing, making similarity searches inefficient. You would have to manage complex indexing logic externally, which is error-prone and can introduce significant latency.
- C. Correct.
C) Correct. Oracle Database supports a vector data type and specialized approximate nearest neighbor indexing. This allows you to insert and update vectors efficiently while enabling fast similarity lookups for semantic search or recommendation use cases.
- D. Incorrect.
D) Incorrect. A brute-force approach in a caching layer (e.g., storing all vectors in memory and scanning them sequentially) can be viable only for small datasets. As data grows, this becomes infeasible and costly in terms of performance and memory usage.