Q64 — AWS SAA-C03 Ch.14

Question 64 of 100 | ← Chapter 14

Q1064. A company is planning to run an AI/ML workload on AWS.The company needs to train a model on a dataset that is in Amazon S3 Standard. A model training application requires multiple compute nodes and single-digit millisecond access to the data.Which solution will meet these requirements in the MOST cost-effective way?

Correct Answer: B. Add partitions to the S3 bucket by adding random prefixes. Reconfigure the model training application to point to the new prefixes as the data source.

Explanation

The solution that meets the company's requirements in the MOST cost-effective way is:B. Add partitions to the S3 bucket by adding random prefixes. Reconfigure the model training application to point to the new prefixes as the data source.Explanation:- Partitioning Data: By adding partitions (e.g., random prefixes), you can optimize the performance of data retrieval in Amazon S3. This can help achieve lower-latency access to the data needed for model training while keeping the data in S3, which is cost-effective compared to other storage options.Other Options:- A. Move the data to S3 Intelligent-Tiering: Although Intelligent-Tiering can optimize costs based on access patterns, it may not provide the low-latency access required for this workload.- C. Move the data to S3 Express One Zone: While this option is cheaper than standard S3, it still does not guarantee the low-latency access required by the application and is not the most cost-effective approach for high-performance workloads.- D. Move the data to a General Purpose SSD (gp3) Amazon EBS volume: This option would provide low- latency access but could be more expensive due to EBS costs, especially for large datasets.Conclusion:Option B is the most cost-effective solution while meeting the requirements for low-latency access during model training.