MLS-C01 Question 13
Select 1You are building a machine learning model to predict customer churn. The dataset contains 500 GB of tabular data that will be accessed frequently during the training phase. Additionally, you must preprocess this dataset and store the preprocessed results for inference. Which storage mediums should you use to store the raw dataset and the preprocessed data, respectively?
- A
Store the raw dataset in Amazon S3 and the preprocessed data in Amazon S3
- B
Store the raw dataset in Amazon S3 and the preprocessed data in Amazon Elastic File System (Amazon EFS)
- C
Store the raw dataset in Amazon Elastic Block Store (Amazon EBS) and the preprocessed data in Amazon S3
- D
Store the raw dataset in Amazon Elastic File System (Amazon EFS) and the preprocessed data in Amazon Elastic File System (Amazon EFS)
- E
Store the raw dataset in Amazon S3 and the preprocessed data in Amazon Elastic Block Store (Amazon EBS)
Show answer and explanation
Correct answer: B
Explanation
Amazon S3 is a cost-effective and scalable choice for storing large datasets that do not require frequent updates or low-latency access, making it ideal for raw datasets. Amazon Elastic File System (Amazon EFS) is a good choice for preprocessed data as it provides low-latency, shared file storage that can be accessed concurrently by multiple training jobs, making it suitable for machine learning workflows. This combination optimizes both cost and performance.
- A. Incorrect.
While you can store both raw and preprocessed data in Amazon S3, this option is not optimal for frequently accessed data during training because Amazon S3 is an object store, not a file system.
- B. Correct.
Storing the raw dataset in Amazon S3 is cost-effective for large datasets, and using Amazon EFS for preprocessed data is ideal due to its low-latency access and file system capabilities, which are suited for machine learning training.
- C. Incorrect.
Amazon EBS is not an ideal storage medium for storing raw datasets, as it is more expensive and better suited for block storage needs. Storing preprocessed data in Amazon S3 is also suboptimal for frequent access during training.
- D. Incorrect.
Storing both raw and preprocessed data in Amazon EFS would work, but using Amazon EFS for the raw dataset is not cost-efficient compared to Amazon S3 for large, infrequently accessed data.
- E. Incorrect.
While storing the raw dataset in Amazon S3 is cost-effective, Amazon EBS is not the best option for preprocessed data, as it is more suitable for block storage needs and less flexible than Amazon EFS for shared file access in training.