Q96 — AWS SAA-C03 Ch.16

Question 96 of 100 | ← Chapter 16

Q1296. A company wants to create a long-term storage solution that will allow users to upload terabytes of images and videos. The company will use the images and videos to train machine learning(ML) models. The storage solution must be scalable and cost-optimized.Which solution will meet these requirements?

Correct Answer: B. Provision an Amazon S3 bucket for users to upload images and videos. Configure the S3 bucket to make the data available to Amazon SageMaker AI for ML model training. Store the data in the S3 Intelligent-Tiering storage class.

Explanation

Let's analyze each option based on the requirements of scalability, cost - optimization, and suitability for ML model training:Option AScalability: Amazon S3 is highly scalable and can easily handle terabytes of image and video uploads. Amazon FSx for Lustre is also scalable, but it is mainly designed for high - performance computing workloads.Cost - optimization: While S3 is cost - effective for long - term storage, Amazon FSx for Lustre is relatively expensive. It is optimized for low - latency, high - throughput access rather than long - term, cost - efficient storage. Using it as an intermediate step for ML training when there are more cost - effective alternatives makes this option less cost - optimized.Suitability for ML training: Although FSx for Lustre can provide fast access to data for ML training, the additional cost and complexity of copying data from S3 to FSx for Lustre are not justified when there are simpler and more cost - effective ways to make S3 data available for ML training.Option BScalability: Amazon S3 is an object storage service that can scale seamlessly to store terabytes or even petabytes of data. It can easily handle the large volume of image and video uploads required by the company.Cost - optimization: The S3 Intelligent - Tiering storage class is designed for long - term data storage with unknown or changing access patterns. It automatically moves data between different access tiers (frequent access, infrequent access, etc.) to optimize costs. This makes it a very cost - effective solution for storing large amounts of data over the long term. Additionally, Amazon SageMaker can directly access data stored in S3, eliminating the need for complex data transfer processes and additional storage infrastructure. Suitability for ML training: Amazon SageMaker is a fully - managed machine learning service that can directly access data stored in S3. This allows for a streamlined workflow where the data stored in S3 can be easily used to train ML models without the need for intermediate data transfers or complex storage setups.Option CScalability: An Amazon SageMaker AI notebook instance with only 16 GB of storage is not scalable enough to handle terabytes of image and video data. It is designed for individual users to develop and test ML models, not for storing large - scale datasets.Cost - optimization: This option is not cost - optimized as it requires users to upload data directly to the notebook instance, which is not an efficient way to store large amounts of data. It also lacks the long - term storage capabilities and cost - effectiveness of S3.Suitability for ML training: While SageMaker notebook instances are useful for ML development, relying on them for large - scale data storage and upload is impractical and not in line with best practices for handling large datasets.Option DScalability: Amazon S3 is scalable for storing large amounts of data, but Amazon EFS is a file system that is more suitable for shared access to files in a network - attached storage (NAS) environment. It is not as cost - effective as S3 for long - term storage of large datasets. Cost - optimization: EFS can be more expensive than S3, especially for long - term storage of large amounts of data. Copying data from S3 to EFS adds an unnecessary step and cost to the process, making this option less cost - optimized.Suitability for ML training: Although Amazon SageMaker can access data from various sources, using EFS as an intermediate step for ML training when S3 can be directly accessed by SageMaker is an over - complication and not the most efficient or cost - effective approach. Based on the above analysis, the solution that meets the requirements of scalability, cost - optimization, and suitability for ML model training is to provision an Amazon S3 bucket for users to upload images and videos, configure the S3 bucket to make the data available to Amazon SageMaker AI for ML model training, and store the data in the S3 Intelligent - Tiering storage class.So, the answer is B.