A machine learning (ML) company continuously develops new versions of its models. The models are integrated into a web application that uses a RESTful API to call custom Python code to invoke the models. The application runs in an Amazon EC2 Auto Scaling group.
Customer feedback indicates that new model versions sometimes perform worse than previous versions. The company wants to implement a gradual deployment approach in which 10% of traffic is initially routed to new models to verify performance. The solution must also reduce operational overhead for infrastructure management.
Which solution will meet these requirements with the LEAST operational overhead?
Submit