1Z0-184-25 Question 54
Single answerYour data science team is configuring an HNSW Vector Index in Oracle Cloud Infrastructure for a product recommendation system. New products are frequently added to your database, and you notice that rebuilding the index after each addition is time-consuming. Which approach is recommended to maintain search accuracy and performance when new data is frequently added?
- A
Recreate the HNSW index from scratch every time a new product is added to ensure consistency
- B
Configure partial index updates for each incremental addition using dynamic concurrency controls
- C
Build the updated index in parallel with new data and swap it with the existing index after validation
- D
Rely on standard SQL queries to supplement HNSW index searches when new data is added
Show answer and explanation
Correct answer: C
Explanation
In Oracle Cloud Infrastructure, HNSW indexes are designed for high-performance vector similarity queries, but they do not efficiently handle frequent incremental updates in-place. The recommended approach, as documented in best practices for approximate nearest neighbor indexing, is to build a new index in parallel whenever there is a significant batch of new data. Once the new index has been validated for accuracy, it is swapped in for production use, ensuring minimal downtime and consistent query performance.
- A. Incorrect.
Option 1 is incorrect because rebuilding the entire index after each addition is inefficient and can lead to significant downtime, especially with a rapidly growing dataset.
- B. Incorrect.
Option 2 is incorrect because HNSW indexes typically do not support direct, fully automated partial updates at scale; attempting concurrency-based partial updates can be complex and may degrade performance.
- C. Correct.
Option 3 is correct. Building a new index in parallel with the updated data, then swapping out the old index once validation is complete, is a recommended best practice as it ensures continuous availability and maintains search quality.
- D. Incorrect.
Option 4 is incorrect because relying on standard SQL queries for new data dilutes the benefits of vector-based similarity searching and can produce inconsistencies, reducing the overall effectiveness of the recommendation system.