1Z0-184-25 Question 96
Single answerYour e-commerce company wants to implement a real-time product recommendation engine using vector embeddings on Oracle Cloud Infrastructure. You plan to generate embeddings for product descriptions and perform sub-second similarity searches to suggest the most relevant products. The solution must scale to millions of embeddings and integrate with existing MySQL-based applications. Which approach best meets these requirements?
- A
Store the vector embeddings in MySQL HeatWave with specialized vector indexes for low-latency similarity queries.
- B
Keep vector embeddings in OCI Object Storage as JSON files and parse them in real time using standard search APIs.
- C
Use Oracle Functions to create embeddings on demand and store them in ephemeral memory for immediate querying.
- D
Insert embeddings in an Autonomous JSON Database table without specialized indexes and rely on standard filters for lookups.
Show answer and explanation
Correct answer: A
Explanation
For low-latency, large-scale vector similarity searches in OCI, specialized vector indexes are crucial. MySQL HeatWave provides native support for vector data types and indexing, enabling sub-second queries on millions of embeddings while integrating directly with MySQL-based applications. Refer to Oracle� documentation on HeatWave Vector Search for best practices and performance guidelines.
- A. Correct.
Correct. MySQL HeatWave supports vector data types and specialized indexing, which helps achieve low-latency similarity searches at scale. This approach also integrates seamlessly with MySQL-based applications, making it suitable for real-time recommendation engines.
- B. Incorrect.
Incorrect. While Object Storage can hold large datasets, repeatedly parsing and searching JSON files for embeddings is inefficient and does not offer real-time vector indexing or sub-second latency capabilities.
- C. Incorrect.
Incorrect. Oracle Functions are well-suited to stateless on-demand processing, not long-term or high-throughput storage of vector embeddings. Storing embeddings only in ephemeral memory during function execution would make it difficult to handle large-scale, continuous recommendation queries.
- D. Incorrect.
Incorrect. Simply inserting embeddings into an Autonomous JSON Database table without specialized indexes would require filtering through large amounts of data, leading to slow searches. Standard secondary indexes are not optimized for vector similarity queries.