QuestionQ22

VerifiedDeployment and Orchestration of ML Workflows

A company uses 10 Reserved Instances of accelerated instance types to serve the current version of an ML model. An ML engineer must deploy a new version of the model to an Amazon SageMaker real-time inference endpoint.

The solution must use the original 10 instances to serve both model versions. It must also include one additional Reserved Instance available for the deployment process. The transition between versions must occur without downtime or service interruptions.

Which solution meets these requirements?

Explanation

A SageMaker rolling deployment updates an endpoint fleet in batches. With a rolling batch size of 1, it provisions one instance for the new model, shifts traffic while the remaining old-model instances continue serving, and then terminates one old instance before the next batch. This requires only one instance of additional capacity during the update and preserves endpoint availability. Blue/green deployments instead provision a full new fleet.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!