Cost Optimization — AWS Machine Learning (MLS-C01)
Utilization pattern determines commitment tier, not instance size.
Savings Plans, Reserved Instances, Spot, and Cost Explorer each address a different dimension of the cost problem. The exam will give you a workload profile — steady-state training, bursty inference, exploratory notebooks — and expect you to match commitment level to utilization certainty. Spot absorbs interruptible training. Savings Plans cover predictable inference capacity. Cost Explorer diagnoses; it doesn't reduce spend. Map the tool to the utilization shape before selecting.
What This Pattern Tests
The exam presents a running workload and asks you to reduce costs. Spot Instances save 60-90% but require fault tolerance. Savings Plans save 20-40% with 1-year or 3-year commitment to a consistent compute spend (flexible across instance families with Compute Savings Plans). Reserved Instances save similarly but lock to specific instance types and regions. Graviton (ARM) instances offer ~20% better price-performance than x86. The trap is recommending Spot for a latency-sensitive web API (interruptions cause errors) or Reserved Instances for a workload that runs 2 hours per day (break-even requires ~40% utilization).
Decision Axis
Workload characteristics (fault tolerance, utilization pattern, flexibility needs) determine which pricing model applies — not just the discount percentage.
Associated Traps
Decision Rules
Whether to replace a custom SageMaker ASR training pipeline with Amazon Transcribe for a commodity speech-to-text workload given a hard 40% cost-per-inference reduction mandate — i.e., apply the build-vs-buy threshold governing constraint.
Whether to replace a custom SageMaker-hosted ASR model with Amazon Transcribe for a commodity standard-English speech-to-text workload, given that the cost-per-inference reduction mandate and variable call volume make the always-on custom endpoint unjustifiable.
When the speech recognition task covers standard language and does not require a domain-differentiated model, replace the custom SageMaker ASR endpoint with Amazon Transcribe batch transcription to eliminate always-on instance-hour billing, data-labeling cost, and retraining overhead — defaulting to the managed AI service wherever the commodity threshold is met.
Whether to use Amazon Comprehend's managed API for commodity sentiment and key-phrase extraction or build and host a custom SageMaker NLP model, given the cost-per-inference mandate, standard English text, and zero dedicated ML operations capacity.
Whether to right-size or reserve the provisioned SageMaker endpoint versus replacing the custom TTS model with Amazon Polly's pay-per-character managed service, eliminating idle capacity waste across twenty hours per day and removing ongoing model maintenance costs.
Domain Coverage
Difficulty Breakdown