MLA-C01 · Question #37
A company has trained an ML model in Amazon SageMaker. The company needs to host the model to provide inferences in a production environment. The model must be highly available and must respond with…
The correct answer is A. Create a SageMaker real-time inference endpoint. Configure auto scaling. Configure the endpoint. Option A is correct because SageMaker real-time inference endpoints are purpose-built for exactly these requirements: they are managed, highly available, low-latency, and natively support auto scaling policies that respond to changing request volumes - making them ideal for…
Question
A company has trained an ML model in Amazon SageMaker. The company needs to host the model to provide inferences in a production environment. The model must be highly available and must respond with minimum latency. The size of each request will be between 1 KB and 3 MB. The model will receive unpredictable bursts of requests during the day. The inferences must adapt proportionally to the changes in demand. How should the company deploy the model into production to meet these requirements?
Options
- ACreate a SageMaker real-time inference endpoint. Configure auto scaling. Configure the endpoint
- BDeploy the model on an Amazon Elastic Container Service (Amazon ECS) cluster. Use ECS
- CInstall SageMaker Operator on an Amazon Elastic Kubernetes Service (Amazon EKS) cluster.
- DUse Spot Instances with a Spot Fleet behind an Application Load Balancer (ALB) for inferences.
How the community answered
(65 responses)- A82% (53)
- B3% (2)
- C9% (6)
- D6% (4)
Explanation
Option A is correct because SageMaker real-time inference endpoints are purpose-built for exactly these requirements: they are managed, highly available, low-latency, and natively support auto scaling policies that respond to changing request volumes - making them ideal for unpredictable burst traffic with request sizes up to several MB.
Why the distractors fail:
- B (ECS): While ECS can host models, it requires significant custom infrastructure work and doesn't natively integrate with SageMaker's model artifacts, auto scaling triggers, or inference optimizations - adding complexity without benefit.
- C (EKS with SageMaker Operator): This is a valid pattern for Kubernetes-native shops, but it is far more operationally complex than a managed endpoint and is not the AWS-recommended path when SageMaker already provides a fully managed solution.
- D (Spot Instances + ALB): Spot Instances can be interrupted at any time, directly violating the high availability requirement. They are cost-optimized for fault-tolerant batch workloads, not latency-sensitive production inference.
Memory tip: Match "SageMaker trained → SageMaker hosted." Whenever a question mentions a model trained in SageMaker that needs low latency, high availability, and scaling, SageMaker real-time endpoints with auto scaling is almost always the answer - it's the fully managed, end-to-end AWS-native path.
Topics
Community Discussion
No community discussion yet for this question.