An ML engineer needs to implement a solution to host a trained ML model. The rate of requests to the model will be inconsistent throughout the day. The ML engineer needs a scalable solution that minim

Question

Options

• ACreate AWS Lambda functions that have fixed concurrency to host the model. Configure the
• BDeploy the model on an Amazon Elastic Container Service (Amazon ECS) cluster that uses AWS
• CDeploy the model to an Amazon SageMaker endpoint. Deploy multiple copies of the model to the
• DDeploy the model to an Amazon SageMaker endpoint. Create SageMaker endpoint auto scaling

• A
  21% (6)
• B
  7% (2)
• C
  10% (3)
• D
  62% (18)

Explanation

Option D is correct because SageMaker endpoint auto scaling dynamically adjusts the number of instances serving the model based on real-time traffic. It scales out to handle peak demand and scales back in during low-traffic periods, directly satisfying both the cost-minimization and peak-capacity requirements.

Option A is wrong because Lambda's fixed concurrency defeats the purpose - fixed concurrency means you're paying for reserved capacity even when idle, and Lambda also has cold-start latency and memory limits that make it poorly suited for large ML models.

Option C is wrong because deploying multiple copies of the model statically (without auto scaling) is the opposite of cost-efficient - you're paying for all those instances 24/7 regardless of traffic.

Option B is wrong because ECS alone (the choice appears truncated, but without an auto-scaling policy configured) doesn't inherently minimize costs during idle periods; it requires additional configuration to match what SageMaker's built-in auto scaling provides out of the box.

Memory tip: Think of it as "scale to need, not to max" - whenever a question mentions inconsistent traffic + minimize idle costs + handle peak load, the answer almost always involves auto scaling (not fixed capacity). On SageMaker, that's endpoint auto scaling; for serverless workloads, SageMaker Serverless Inference is another variant worth knowing.

No community discussion yet for this question.