mediumMultiple Select
MLA-C01 Practice Question: A team is training a deep learning model on…
A team is training a deep learning model on Amazon SageMaker using a custom Docker container. Which three practices should they follow to optimize training performance? (Choose three.)
⚠ Common exam trap
AWS often tests the misconception that bigger instances always mean faster training, but the real optimization lies in data pipeline efficiency (e.g., Pipe mode, compression, and shuffling) and cost management (e.g., Managed Spot Training with checkpointing).
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Store training data in Amazon S3 in a shuffled and compressed format
Option A is correct because storing training data in Amazon S3 in a shuffled and compressed format reduces storage footprint and download time, and shuffling prevents order-dependent bias while improving I/O throughput during training. Option D is correct because SageMaker Managed Spot Training with checkpointing lets the team use discounted spare EC2 capacity and resume from the last checkpoint after an interruption, cutting cost without losing training progress. Option E is correct because Pipe mode streams data directly from S3 to the training container, avoiding the latency and disk overhead of downloading the full dataset as File mode does, which speeds up training for large datasets. Option B is not necessarily right because the largest instance type may be cost-inefficient or mismatched to the workload, and performance depends on proper resource sizing rather than simply maximizing instance size. Option C is not right because adding layers increases model complexity and training time and does not by itself optimize training performance.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✓
Store training data in Amazon S3 in a shuffled and compressed format
Why this is correct
Storing data shuffled and compressed in Amazon S3 reduces per-epoch download time and prevents order-induced bias during training. Compression lowers I/O volume, while shuffling ensures the data loading pipeline feeds batches without sequential correlation, directly improving throughput for the custom container's training job.
- ✗
Use the largest instance type available
Why it's wrong here
Larger instances add cost and do not remove bottlenecks such as data-loading or GPU underutilisation; the correct practices are distributed training, efficient input pipelines and profiling. It is tempting because more vCPUs and memory can help, but only when compute is genuinely the limiting factor.
- ✗
Increase the number of layers in the model to improve accuracy
Why it's wrong here
Adding layers increases parameter count and training time without guaranteeing accuracy, so it does not optimise performance. It is tempting because deeper networks can model complex patterns, but the required practices are mixed precision, distributed training and optimised input pipelines.
- ✓
Use SageMaker Managed Spot Training with checkpointing
Why this is correct
Managed Spot Training uses spare EC2 capacity at reduced cost, and checkpointing writes model state to Amazon S3 so interrupted jobs resume rather than restart. This satisfies the requirement to optimise training performance without losing progress on interruption.
- ✓
Use Pipe mode to stream data instead of File mode
Why this is correct
Pipe mode streams training data directly from Amazon S3 into the container, avoiding the full dataset download and disk write that File mode performs first. This removes the startup latency and storage overhead that constrain large training datasets.
Quick reference
AWS S3 Storage Class Comparison
| Storage Class | Min Duration | Retrieval | Use Case |
|---|---|---|---|
| S3 Standard | None | Immediate | Frequently accessed data |
| S3 Standard-IA | 30 days | Immediate | Infrequent access, rapid retrieval |
| S3 One Zone-IA | 30 days | Immediate | Non-critical infrequent data |
| S3 Intelligent-Tiering | None | Immediate–hours | Unknown or changing access patterns |
| S3 Glacier Instant | 90 days | Milliseconds | Archive with instant retrieval |
| S3 Glacier Flexible | 90 days | Minutes–hours | Archive, flexible retrieval |
| S3 Glacier Deep Archive | 180 days | Hours | Long-term compliance archive |
Go deeper
Related to this question
About these practice questions
One of 665 original MLA-C01 practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This MLA-C01 practice question is part of Courseiva's free Amazon Web Services certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the MLA-C01 exam.