An inference system must run a feature-normalization container followed by a classifier container, so each request passes through both in order behind one endpoint. Which SageMaker deployment option matches this?
Show answer and explanation
Answer: A
A SageMaker inference pipeline, also called a serial multi-container endpoint, chains up to 15 containers behind one endpoint. Each request flows through the containers in sequence, so preprocessing and the model run in a single endpoint call. Multi-model endpoints host many independent models, serverless endpoints scale on demand without chaining, and Greengrass deployment targets devices outside the cloud.
Why the other options are wrong
- B. Multi-model endpoints share one serving stack and serve many independent models, loading them on demand. Requests are not passed through a sequence of containers.
- C. Serverless inference scales automatically for intermittent traffic. It does not chain preprocessing and inference containers in sequence for each request.
- D. Greengrass deployment delivers components and models to edge devices. It is not a cloud endpoint that chains containers for each request.