An advertising company stores terabytes of data in an Amazon S3 data lake. The company wants to build its own foundation model (FM) and has deployed a training cluster on AWS. The company loads…

Question

Options

• AMount an Amazon Elastic File System (Amazon EFS) file system to the training cluster. Use AWS
• BUse an Amazon FSx for Lustre file system and Amazon S3 with Data Repository Association
• CAttach Amazon Block Store (Amazon EBS) volumes to the training cluster. Load the data from
• DUse AWS DataSync to migrate the data from Amazon S3 to the training cluster as files. Train the

• A
  33% (15)
• B
  43% (20)
• C
  9% (4)
• D
  15% (7)

Explanation

Amazon FSx for Lustre is a high-performance parallel file system designed for machine learning and HPC that can be linked to Amazon S3 via Data Repository Associations (DRA). With DRA, you can import S3 objects as files (lazy or preloaded) and export results back to S3, providing very high throughput and low-latency POSIX access on the training cluster. This S3-native integration minimizes data loading overhead and accelerates training cycles at scale. EFS (A) is general-purpose and lower throughput per TB than Lustre. EBS (C) requires manual copy and does not scale as a shared parallel filesystem. DataSync alone (D) is a transfer tool, not a high- throughput training filesystem. FSx for Lustre with S3 DRA best satisfies scalability, throughput, and native S3 integration for rapid ML training.

No community discussion yet for this question.