A recommendation model uses ML and calls an Amazon SageMaker AI endpoint to get recommendations. An ML engineer must ensure that the model stays available during an expected increase in user traffic.

Question

Options

• AConfigure auto scaling on the SageMaker AI endpoint.
• BCreate a new SageMaker AI endpoint. Deploy the model to the new endpoint.
• CUse SageMaker Neo to optimize the model for inference.
• DAttach an Auto Scaling group to the SageMaker AI endpoint.

• A
  70% (16)
• B
  4% (1)
• C
  17% (4)
• D
  9% (2)

Explanation

Configuring auto scaling on the SageMaker AI endpoint allows it to automatically add or remove inference instances in response to changes in traffic load. You define a scaling policy (e.g., based on InvocationsPerInstance CloudWatch metric), and SageMaker scales the endpoint's instance count within defined minimum and maximum bounds-ensuring availability during traffic spikes without over-provisioning during quiet periods. Option B creates an additional endpoint but does not scale the existing one to handle the increased load dynamically. Option C (SageMaker Neo) optimizes the model's inference speed and resource efficiency but does not add capacity to handle more concurrent requests. Option D is incorrect because SageMaker endpoints use Application Auto Scaling (managed by SageMaker), not EC2 Auto Scaling groups, which attach to EC2 instances directly.

No community discussion yet for this question.