A public-works team needs containerized model inference that requires GPU instance types. Managed Rekognition does not cover their custom model. Which compute placement meets the special-hardware need?
Select an answer to reveal the explanation.
Short Explanation
Custom models that need GPUs can’t hide in tiny Lambda or a cache node. Run the containers on ECS backed by EC2 GPU instance types when Rekognition isn’t a fit. Transfer Acceleration speeds uploads; it doesn’t add GPUs, and ElastiCache isn’t an inference accelerator.
Full Explanation
When workloads require specialized hardware such as GPUs for custom model inference, ECS on EC2 lets you select instance types that expose those features. Purpose-built services like Amazon Rekognition should be preferred when they meet requirements, but DIY GPU containers need EC2-backed capacity. Lambda default memory, Transfer Acceleration, and ElastiCache do not provide GPU inference hosting.