1Z0-184-25 Question 53
Select 2You manage an e-commerce platform running on Oracle Autonomous Database with a large catalog of embedded product descriptions. You use an HNSW vector index to speed up semantic similarity searches, but as more items are added, your query latency has increased and recall accuracy sometimes dips. Which two actions should you take to optimize the HNSW index performance and maintain accurate search results?
- A
Periodically rebuild or reindex segments after bulk inserts to keep vector data structures organized
- B
Increase the efSearch parameter to ensure enough candidates are explored during query time
- C
Lower the M parameter substantially to reduce index size and speed up inserts
- D
Enable automatic query caching at the database level by setting the HNSW_CACHE parameter
- E
Use parallel execution to handle concurrency, while balancing resources to avoid index contention
Show answer and explanation
Correct answers: A, B
Explanation
HNSW (Hierarchical Navigable Small World) is designed for approximate nearest neighbor searches over high-dimensional vector embeddings. Managing index quality and tuning search parameters are key to balancing performance and accuracy. Inserting large batches of new products can fragment the index, so periodic refresh or reindexing helps maintain efficiency. Additionally, adjusting HNSW parameters such as efSearch improves recall through expanded candidate exploration. Refer to Oracle best practices on vector indexing in Autonomous Database for guidance on tuning M, efSearch, and reindexing strategies.
- A. Correct.
Correct. Regularly rebuilding or reindexing ensures the HNSW graph remains efficient despite data growth. Bulk insert operations can fragment the index structures, so periodic refresh helps maintain consistent performance.
- B. Correct.
Correct. The efSearch parameter determines how many candidate vectors are explored at query time. Increasing it can improve result accuracy (recall), though it may slightly increase latency. Proper tuning balances performance and accuracy.
- C. Incorrect.
Incorrect. Lowering the M parameter drastically reduces graph connectivity. While it might reduce index size, it can also hurt similarity accuracy and recall. Overly small M results in lower-quality neighbor connections, decreasing result correctness.
- D. Incorrect.
Incorrect. There is no dedicated parameter called HNSW_CACHE in Oracle Autonomous Database for automatically caching vector queries. Caching might be implemented at the application or query layer, but it is not a standard HNSW index parameter.
- E. Incorrect.
Incorrect. Although parallel execution can improve query throughput, HNSW indexing performance must first be addressed by ensuring the graph and query parameters (like efSearch and M) are tuned properly. Parallel execution alone does not resolve underlying index structure or parameter issues.