As an AWS Machine Learning Engineer, you need to deploy a fraud detection model for a financial institution using Amazon SageMaker. The model must quickly adapt to emerging fraud patterns, and updates must be rolled out safely without causing downtime or performance degradation. Which deployment strategy should you use to ensure high availability and a reliable update mechanism?