A company has an ML model that is deployed to an Amazon SageMaker endpoint for real-time inference. The company needs to deploy a new model. The company must compare the new model's performance to…

Question

Options

• ADeploy the new model to a separate endpoint. Manually split traffic between the two endpoints.
• BDeploy the new model to a separate endpoint. Use Amazon CloudFront to distribute traffic
• CDeploy the new model as a shadow variant on the same endpoint as the current model. Route a
• DUse AWS Lambda functions with custom logic to route traffic between the current model and the

• A
  17% (6)
• B
  3% (1)
• C
  74% (26)
• D
  6% (2)

Explanation

Option C is correct because SageMaker natively supports shadow variants on a single endpoint, allowing you to mirror a copy of live inference traffic to the new model for performance comparison - all without changing the response returned to users or managing separate infrastructure. This built-in feature requires minimal configuration and no custom code.

Why the distractors fail:

• A requires provisioning and managing two separate endpoints and manually engineering traffic splitting logic - significantly more operational overhead.
• B introduces Amazon CloudFront, which is a CDN designed for content delivery, not ML model traffic routing - wrong tool for the job and adds unnecessary complexity.
• D requires writing, deploying, and maintaining custom Lambda routing logic, which is the highest operational effort of all options.

Memory tip: On AWS exams, "least operational effort" almost always points to a native, managed feature of the service already in use. If SageMaker can handle it internally (one endpoint, built-in shadow testing), you never need an external service or custom code.

No community discussion yet for this question.