AWS Certified Machine Learning – SpecialtyMachine Learning Implementation and OperationsMedium
A manufacturing company uses a machine learning model to predict equipment failures. The model is deployed on an Amazon SageMaker endpoint. To reduce inference costs, the company wants to utilize specialized hardware accelerators to improve throughput and reduce latency without provisioning full GPU instances, as the model's complexity does not warrant it. Which SageMaker feature allows attaching fractional GPU power to CPU instances for cost-effective acceleration?
- ASageMaker Inference Recommender
- BSageMaker Multi-Model Endpoints
- CSageMaker Neo
- DSageMaker Elastic Inference
Show answer & explanationAnswer & explanation
Correct answer: D. SageMaker Elastic Inference
SageMaker Elastic Inference allows you to attach fractional GPU acceleration to any SageMaker CPU instance, providing a cost-effective way to boost inference performance for deep learning models without the overhead of full GPU instances.
Why the other options are wrong
- A. SageMaker Inference Recommender helps choose the best instance type and configuration for inference, but it's a recommendation tool, not an acceleration feature itself.
- B. SageMaker Multi-Model Endpoints host multiple models on a single endpoint, reducing operational overhead, but do not provide fractional GPU acceleration.
- C. SageMaker Neo compiles models for various hardware platforms to optimize performance, but it doesn't provide fractional GPU acceleration for existing CPU instances.
SageMaker Elastic Inference
Amazon SageMaker Elastic Inference allows you to attach configurable amounts of GPU-powered inference acceleration to Amazon EC2 and SageMaker instances, reducing inference costs.
- Attaches fractional GPU acceleration to CPU instances.
- Cost-effective for deep learning inference.
- Improves throughput and reduces latency.
Memory trick: Elastic Inference stretches your budget by adding just enough power.