A company wants to use Amazon SageMaker to host an ML model that runs on CPU for real-time predictions. The model will have intermittent traffic during business hours and will have periods of no…

Question

Options

• ADeploy the model to a SageMaker real-time endpoint. Add a schedule-based auto scaling policy
• BDeploy the model to a SageMaker Serverless Inference endpoint. Configure increased
• CDeploy the model to a SageMaker Asynchronous Inference endpoint. Configure an auto scaling
• DDeploy the model to a SageMaker real-time endpoint. Create a scheduled AWS Lambda function

• A
  6% (2)
• B
  76% (25)
• C
  3% (1)
• D
  15% (5)

Explanation

SageMaker Serverless Inference is purpose-built for intermittent, unpredictable traffic - it scales to zero during idle periods (after business hours) so you pay only for the compute consumed per request, making it the most cost-effective choice for this workload profile.

Why the distractors fail:

• A - A real-time endpoint keeps instances running 24/7; even with schedule-based auto scaling, you're still paying for idle capacity during off-hours, not eliminating it.
• C - Asynchronous Inference is designed for long-running, queue-based jobs (large payloads, batch-like workloads), not low-latency real-time predictions, and it doesn't natively scale to zero between requests.
• D - A scheduled Lambda to manage a real-time endpoint adds operational complexity and still can't fully eliminate costs during the gaps between scale-down and scale-up windows.

Memory tip: Think "Serverless = pay-per-request + zero idle cost." Whenever an exam question pairs intermittent traffic + no traffic periods + cost optimization, Serverless Inference is almost always the answer - it's the only SageMaker hosting option that truly scales to zero.

No community discussion yet for this question.