MLA-C01 Deployment and Orchestration of ML Workflows Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Asynchronous Inference

  Why this is correct

  Asynchronous Inference queues requests and supports payloads up to 1 GB, writing results to Amazon S3 and publishing completion notifications via Amazon SNS. This directly satisfies the stem's large-payload, asynchronous and notification constraints, unlike real-time endpoints, which cap payloads far smaller.

• ✗

  Batch Transform

  Why it's wrong here

  Batch Transform writes predictions to S3 but provides no built-in SNS completion notification, and it is designed for offline jobs over stored datasets rather than per-request asynchronous invocation. It is tempting because it handles large volumes to S3, and it would be correct for scheduled bulk scoring without notification requirements.

• ✗

  Real-time endpoint

  Why it's wrong here

  Real-time endpoints cap request payloads at roughly 6 MB and return results synchronously, so 1 GB payloads cannot be submitted. It is tempting because endpoints are the default deployment target, and a real-time endpoint would be correct for low-latency interactive predictions on small inputs.

• ✗

  Serverless Inference

  Why it's wrong here

  Serverless Inference enforces a much smaller payload limit and a 60-second timeout, so 1 GB payloads cannot be processed. It is tempting because it removes infrastructure management and scales to zero, and it would be correct for sporadic, small, latency-tolerant inference requests.

Go deeper

Related to this question