1Z0-184-25 Question 100
Single answerYou are designing an advanced semantic search system that stores documents in an Autonomous Database. To handle large-scale text data, you want to generate vector embeddings outside the database using GPU-based inference in Oracle Cloud Infrastructure (OCI) Data Science and then load these embeddings back into the Autonomous Database for downstream analytic queries. Which approach is the most secure and cost-effective way to accomplish this at scale?
- A
Export the documents to Oracle Cloud Infrastructure Object Storage, process them with an OCI Data Science notebook session for embeddings, and load the vectors into your Autonomous Database using Oracle Data Integration or bulk load tools.
- B
Use a custom PL/SQL procedure inside the Autonomous Database to call out via REST APIs to an external GPU server, returning the embeddings directly to the database.
- C
Deploy an on-premises GPU cluster and transfer the embeddings to the Autonomous Database using a public endpoint for quick data ingestion.
- D
Set up an on-demand Oracle Functions application to retrieve documents from the database, generate embeddings using CPU-based processing, and reinsert them via a synchronous call.
Show answer and explanation
Correct answer: A
Explanation
Generating vector embeddings for large datasets outside the Oracle database often involves using Oracle Cloud Infrastructure Data Science, which offers GPU-powered compute resources tailored for machine learning workloads. By temporarily storing and staging input and output data in Oracle Cloud Infrastructure Object Storage, organizations can securely and efficiently pipe embeddings into an Autonomous Database. Oracle Data Integration or similar bulk-loading tools provide a streamlined method to ingest these vectors back into the database, following OCI best practices for scalability and cost management. See Oracle� documentation on 'Oracle Cloud Infrastructure Data Science' and 'Data Integration best practices' for further guidance.
- A. Correct.
Correct. Storing source text in OCI Object Storage, generating embeddings with an OCI Data Science notebook (leveraging GPU shapes), and then using Oracle Data Integration or a bulk load method to insert embeddings into the Autonomous Database is a recommended pattern. It balances security, scalability, and cost efficiency. Object Storage becomes the staging area, and Data Integration or SQL*Loader-type tools handle large-scale data ingestion. This approach is outlined in many OCI reference architectures.
- B. Incorrect.
Incorrect. While it might seem convenient to have the database call an external GPU service via PL/SQL, this can introduce complex network configurations, potential security risks, and inefficient data transfers. Generating embeddings at scale directly through PL/SQL calls to an external service isn�t generally recommended for large datasets.
- C. Incorrect.
Incorrect. Using an on-premises GPU cluster to generate embeddings could be feasible, but relying on a public endpoint for data transfers can raise security and performance concerns. OCI provides more secure private connectivity and integrated services that reduce overhead and complexity.
- D. Incorrect.
Incorrect. Oracle Functions support short-lived serverless executions. However, generating embeddings at scale using CPU-based processing would be inefficient and costly for large volumes of data, and constant synchronous calls could lead to operational bottlenecks.