1Z0-184-25 Question 159
Single answerYou�re designing a recommendation engine that relies on millions of product vector embeddings stored in an Oracle Autonomous Database deployed on Exadata with AI Storage. To achieve minimal data movement and maximum performance during vector similarity searches, which strategy should you implement?
- A
Store all vector data in traditional row-based tables, rely on standard B-tree indexes, and expect the database optimizer to handle vector queries effectively.
- B
Retrieve vector embeddings from external object storage on-demand to reduce Exadata storage use and mitigate data processing overhead.
- C
Maintain specialized vector indexes within Exadata AI Storage, leveraging the storage layer� local compute and memory resources for high-performance similarity searches.
- D
Distribute vector data evenly across application server disks outside of Exadata, reducing dependence on propriety Exadata features.
Show answer and explanation
Correct answer: C
Explanation
When using Exadata to accelerate AI vector searches, placing vector data in specialized indexes on Exadata AI Storage is key to minimizing data movement and leveraging specialized hardware accelerations for vector operations. According to Oracle� best practices, replacing row-based indexing with specialized vector indexes significantly improves the performance of similarity searches on large embeddings. Refer to the latest Oracle Exadata documentation and Oracle Database 23c features for guidance on implementing these vector indexing capabilities.
- A. Incorrect.
Option 1 is incorrect. B-tree indexes are not designed for vector-based similarity searches. Storing embeddings in row-based tables and relying on traditional indexing typically leads to slow query performance and higher data movement, making it an unsuitable approach for large-scale vector workloads.
- B. Incorrect.
Option 2 is incorrect. While using external object storage can save space on Exadata, it does not reduce data movement overhead for AI vector searches. You would incur frequent data transfers from external storage, negatively impacting latency and performance.
- C. Correct.
Option 3 is correct. Exadata AI Storage can handle specialized vector indexes closer to the data, reducing data movement between compute and storage nodes and offering hardware accelerations for vector operations. This substantially improves the performance of similarity searches on large embeddings.
- D. Incorrect.
Option 4 is incorrect. Moving vector data to application server disks negates the benefits of Exadata� optimized storage layer and specialized vector indexing capabilities. This approach disperses data and often increases query latency due to non-optimized storage and compute.