1Z0-184-25 Question 97
Single answerYour analytics team is building a recommendation system in Oracle Cloud Infrastructure for a large dataset of product descriptions. They have generated vector embeddings that capture semantic relationships between products. The requirement is to enable real-time queries returning semantically similar items, achieving millisecond-level response times for end users. Which storage and indexing strategy should you recommend in Oracle Cloud Infrastructure to meet these performance goals?
- A
Store embeddings in Cloud Object Storage as flat files, and rely on an external vector search library maintained by the application.
- B
Use a specialized vector data type in Oracle Autonomous Database and create an approximate nearest neighbor (ANN) index for vector similarity queries.
- C
Create a standard B-tree index on the numerical values of the embeddings in Oracle Autonomous Database.
- D
Import the embeddings into an Oracle NoSQL Database table and query them using primary key lookups for best performance.
Show answer and explanation
Correct answer: B
Explanation
In Oracle Cloud Infrastructure, leveraging the specialized vector data type and approximate nearest neighbor (ANN) capabilities in Oracle Autonomous Database is the most effective way to achieve low-latency vector similarity queries. By using the integrated vector index, you can perform semantic searches across large datasets with near real-time responsiveness. Refer to Oracle documentation on Autonomous Database Vector Search for detailed guidance on setup, indexing strategies, and performance tuning.
- A. Incorrect.
Option 1 is incorrect. While you can store data in Cloud Object Storage, externalizing the search logic and relying on flat files typically leads to higher latency and adds significant operational complexity. This approach does not leverage optimized, built-in vector search capabilities, which are critical for real-time performance.
- B. Correct.
Option 2 is correct. Storing embeddings in Oracle Autonomous Database using a specialized vector data type and creating an approximate nearest neighbor (ANN) index enables fast similarity queries at scale. This strategy uses built-in functionality optimized for vector operations and allows you to leverage database performance and concurrency features in Oracle Cloud Infrastructure.
- C. Incorrect.
Option 3 is incorrect because a standard B-tree index is not suitable for high-dimensional vector data. B-tree indexes work well for scalar comparisons, but they are not optimized for measuring semantic similarity using distance metrics commonly found in embeddings.
- D. Incorrect.
Option 4 is incorrect. While Oracle NoSQL Database can handle large volumes of data, querying by primary key does not address the need to retrieve nearest neighbors based on vector similarity. You would still need specialized indexing, which is not addressed by key-based lookups.