You need to deploy a scikit-learn classification model to production. The model must serve requests 24/7, and you expect millions of requests per second to the production application between 8 AM and 7 PM. You need to minimize the cost of deployment. What should you do?
