What is NOT a recommended activity in the solutions design process for an ECS deployment focused on machine learning tasks?

Question

Options

• AIgnoring the importance of data privacy and security
• BBenchmarking different machine learning frameworks
• CSelecting appropriate GPU-accelerated nodes
• DEvaluating different data preprocessing methods

• A
  86% (30)
• B
  3% (1)
• C
  9% (3)
• D
  3% (1)

Explanation

Ignoring data privacy and security (A) is the correct answer because it represents a critical failure in any ML deployment design - not a recommended activity. Security and privacy are non-negotiable requirements, especially when ML workloads often process sensitive datasets.

Why the distractors are wrong (i.e., why B, C, and D ARE recommended):

• B (Benchmarking ML frameworks): Comparing frameworks like TensorFlow vs. PyTorch in your ECS environment ensures you pick the best-performing tool for your specific workload.
• C (Selecting GPU-accelerated nodes): ML training and inference are compute-intensive; choosing the right GPU-backed node types is a core infrastructure decision.
• D (Evaluating data preprocessing methods): Data quality and pipeline efficiency directly impact model accuracy and throughput, making this a standard design-phase activity.

Memory tip: Think of the "NOT recommended" option as the one that would make a security auditor cringe. In any cloud deployment - ML or otherwise - ignoring security is never a valid design choice. If an answer sounds like negligence rather than a tradeoff, it's likely the "NOT recommended" pick.

No community discussion yet for this question.