How should training infrastructure be created repeatably?
Why D is the answer
With infrastructure as code, such as AWS CloudFormation
Infrastructure as code is reviewable and repeatable.
In detail
- Infrastructure as code is repeatable and reviewable.
- Templates can be reviewed like code.
Use case
An illustrative scenario, not a real incident.
A team defines training infrastructure in CloudFormation, so test and production match.
See it in different ways
flowchart LR A["CodeBuild: build the image"] --> B["ECR: store it"] B --> C["Pipelines: train and evaluate"] C --> D["Model Registry: approve"] D --> E["Endpoint"]
flowchart LR S0["Write template"] --> S1["Review"] --> S2["Deploy stack"]
Key point: Infrastructure as code is repeatable.
Why the other options are wrong
- A. By clicking through the console each time
- Console clicks drift.
- B. By copying settings from a chat message
- Chat messages are not a template.
- C. By asking each engineer to make their own
- Each person doing it differently drifts.

