1Z0-184-25 Question 50
Single answerYou are implementing a semantic search solution on Oracle Cloud Infrastructure using an HNSW vector index for 20 million product embeddings in a high-traffic e-commerce application. You need a strategy that balances index build time, memory usage, and incremental data updates to ensure consistent performance. Which approach is recommended?
- A
Build the entire HNSW index once with the full dataset and rebuild it from scratch whenever new product embeddings arrive.
- B
Batch new embeddings periodically for incremental updates while tuning indexing parameters to maintain an optimal balance between accuracy and speed.
- C
Activate ephemeral indexing on each query to generate the HNSW structure in-memory on the fly.
- D
Disable concurrency control during index generation to streamline indexing operations for large datasets.
Show answer and explanation
Correct answer: B
Explanation
Incremental updates are crucial for maintaining performance and scalability when using HNSW indexes on Oracle Cloud Infrastructure. By batching new embeddings and tuning parameters (e.g., M for connectivity and ef for search quality), you ensure both high-quality search results and efficient runtime performance. Refer to official Oracle documentation for best practices on parameter tuning and indexing strategies with large-scale vector data.
- A. Incorrect.
Option 1 is incorrect because rebuilding from scratch repeatedly is expensive and impractical with large-scale datasets, leading to significant downtime or degraded performance.
- B. Correct.
Option 2 is correct because incremental updates in batches are a recommended practice for HNSW indexes, ensuring that the index remains current without incurring the full cost of a complete rebuild each time. Proper tuning of parameters like M and ef helps strike the right balance between accuracy and performance.
- C. Incorrect.
Option 3 is incorrect because generating the index on the fly for each query is highly inefficient and impractical for production workloads with large embeddings. HNSW indexes benefit from being pre-built instead of recalculated per query.
- D. Incorrect.
Option 4 is incorrect because disabling concurrency control can lead to index corruption or inconsistent states under heavy load. Concurrency management is necessary for stable and reliable indexing.