MLS-C01 Question 246
Select 2You are deploying a machine learning model using Amazon SageMaker and need to ensure the deployed model can handle sudden spikes in inference requests while optimizing costs. Which combination of features in SageMaker would best address this requirement?
- A
Enable Amazon SageMaker Multi-Model Endpoint
- B
Use Amazon SageMaker Auto Scaling for the endpoint
- C
Deploy the model on an Amazon SageMaker Serverless Inference endpoint
- D
Manually over-provision multiple instances for the endpoint
- E
Enable Amazon SageMaker Batch Transform for inference requests
Show answer and explanation
Correct answers: B, C
Explanation
To handle sudden spikes in inference requests while optimizing costs, combining SageMaker Auto Scaling and Serverless Inference provides a scalable and cost-effective solution. Auto Scaling dynamically adjusts resources based on demand, and Serverless Inference eliminates the need for provisioning and managing servers, making it ideal for unpredictable workloads.
- A. Incorrect.
Multi-Model Endpoints are used to host multiple models on a single endpoint to save costs, but they are not designed to handle sudden spikes in inference requests.
- B. Correct.
Auto Scaling adjusts the number of instances in response to traffic patterns, ensuring cost optimization and the ability to handle sudden increases in requests.
- C. Correct.
Serverless Inference automatically provisions resources to handle varying workloads, making it suitable for unpredictable traffic patterns while optimizing costs.
- D. Incorrect.
Manually over-provisioning multiple instances can handle spikes but is not cost-efficient as resources are underutilized during normal traffic.
- E. Incorrect.
Batch Transform is used for offline inference on large datasets and is not suitable for real-time, spike-heavy inference scenarios.