Amazon SageMaker Inference now supports rolling update for inference component endpoints

Amazon SageMaker Inference has introduced rolling updates for inference component (IC) endpoints, offering a more efficient way to update machine learning models.

• Rolling updates allow customers to update endpoints batch by batch, without traffic interruption
• Replaces previous blue/green update method that required doubling instances
• Reduces the number of additional instances needed during model updates
• Helps customers minimize costs and capacity reservation requirements
• Available in multiple regions across Asia Pacific, Canada, Europe, Middle East, South America, and US

This feature enables more flexible and cost-effective deployment of machine learning models, particularly foundation models, on SageMaker Inference endpoints.