A company plans to deploy an ML model for production inference on an Amazon SageMaker endpoint. The average inference payload size will vary from 100 MB to 300 MB. Inference requests must be…

Question

Options

• AServerless inference
• BAsynchronous inference
• CReal-time inference
• DBatch transform

• A
  7% (3)
• B
  74% (32)
• C
  14% (6)
• D
  5% (2)

Explanation

Asynchronous inference is purpose-built for exactly these two constraints: it supports payloads up to 1 GB and processing times up to 1 hour, making it the only option that satisfies both the 100–300 MB payload range and the 60-minute SLA.

Why the others fail:

• A. Serverless inference - Supports tiny payloads (max ~4 MB) and is designed for sporadic, lightweight requests; it cannot handle 100–300 MB.
• C. Real-time inference - Has a 6 MB payload limit and expects low-latency (sub-second) responses; both limits are violated here.
• D. Batch transform - Processes large datasets offline from S3 in bulk jobs, not individual production inference requests via an endpoint; it's the wrong paradigm entirely.

Memory tip: Associate Async = A-sync-ually large and slow is fine - when you see big payloads + long processing time + endpoint, that's the async inference signature. The other three options all have hard limits (payload or latency) that rule them out as soon as you see "100–300 MB" in the question.

No community discussion yet for this question.