1Z0-184-25 Question 99
Select 2Your business has a large text dataset stored in an Oracle Autonomous Database for a new recommendation engine. The data science team wants to generate vector embeddings outside the database using Oracle Cloud Infrastructure Data Science. Which two strategies reflect recommended best practices to accomplish this while minimizing overhead and ensuring secure data transfer?
- A
Use an Oracle Cloud Infrastructure Data Science notebook session with a private endpoint to retrieve data directly from the Autonomous Database, generate embeddings, and write them to Object Storage.
- B
Provide direct root-level OS access from the Autonomous Database DB system to the Data Science environment so data can be streamed in real time without further configuration.
- C
Ensure that required security policies and data masking are applied before exporting or streaming data, especially if it contains sensitive information, to comply with organizational and regulatory requirements.
- D
Store all generated embeddings on an attached block volume in the Autonomous Database for immediate in-database availability and direct control.
Show answer and explanation
Correct answers: A, C
Explanation
When generating vector embeddings outside the Oracle Database, you should ensure a secure, direct connection from Oracle Cloud Infrastructure Data Science to your source data. Properly configured private endpoints, service gateways, and security policies enable seamless data retrieval without compromising security. Storing results in Object Storage allows for scalability, ease of integration with other services, and better cost effectiveness. Detailed information on creating secure connections and best practices can be found in the Oracle Cloud Infrastructure Data Science documentation and Oracle Architecture Center reference architectures.
- A. Correct.
Correct. Establishing a secure connection from the Data Science notebook to the Autonomous Database (for instance, using a private endpoint or a service gateway) is a standard and recommended approach. You avoid unnecessary data movement to external solutions by reading data directly. Storing the generated embeddings in Object Storage then makes them easily accessible for subsequent machine learning or analytics processes.
- B. Incorrect.
Incorrect. Granting direct root-level OS access from the DB system to any external environment is a major security concern. Oracle best practices discourage sharing root-level access between services because it can bypass security policies and expose your data and system to risks.
- C. Correct.
Correct. Sensitive data should be masked or protected before it leaves the database environment. Enforcing security policies and data masking helps maintain compliance, reduce risks, and align with standard data governance procedures within Oracle Cloud Infrastructure.
- D. Incorrect.
Incorrect. While you can store some data in block volumes, it is not the recommended pattern for large-scale vector embeddings or for external consumption. Object Storage is more flexible and scalable for sharing model outputs and intermediate datasets between various OCI services and external applications.