1Z0-184-25 Question 126
Single answerYou are creating a Retrieval-Augmented Generation (RAG) solution on Oracle Cloud Infrastructure (OCI) for a large internal knowledge base containing millions of documents stored in Oracle Object Storage. You plan to generate vector embeddings for these documents to enable high-performance retrieval. Which approach best ensures both cost-efficiency and near-real-time query latency for storing and retrieving these embeddings?
- A
Store and index documents in OCI Object Storage, relying on the Large Language Model (LLM) to handle real-time chunking and retrieval.
- B
Deploy a GPU-accelerated instance and store embeddings in its ephemeral block storage, re-generating them at each session.
- C
Use Oracle Autonomous Database with built-in vector search capabilities to store embeddings and retrieve relevant chunks.
- D
Use OCI Logging to store vector embeddings since Logging handles large volumes of data by default.
Show answer and explanation
Correct answer: C
Explanation
When building a RAG application on Oracle Cloud Infrastructure, you typically need a data store that supports dense vector similarity searches and can scale to billions of embeddings. Oracle Autonomous Database with vector search capabilities is an optimal choice. It simplifies the ingestion process, provides low-latency querying, and scales seamlessly. Refer to Oracle documentation on Autonomous Database vector search for details on implementing RAG with Oracle Cloud.
- A. Incorrect.
Option 1: Incorrect. While Object Storage is excellent for persisting large volumes of unstructured data, it does not provide the indexing or query mechanisms needed for sub-second vector similarity searches. Relying on the LLM to handle all chunking in real time can become prohibitively expensive and slow.
- B. Incorrect.
Option 2: Incorrect. Storing embeddings in ephemeral block storage on a GPU instance requires constant re-generation each time the instance restarts and can become costly. This approach also complicates horizontal scaling since state is confined to a single instance.
- C. Correct.
Option 3: Correct. Autonomous Database (including Autonomous JSON Database or Autonomous Data Warehouse) supports vector search capabilities and provides efficient indexing. Storing embeddings here allows you to perform near-real-time similarity search without the overhead of re-indexing each session. It also scales both in storage and compute to handle large data sets.
- D. Incorrect.
Option 4: Incorrect. OCI Logging is designed for log data ingestion and analysis, not for storing and querying vector embeddings. While it can handle large volumes, it does not offer the necessary retrieval performance for RAG applications.