An AWS Elastic Beanstalk application is deployed with an Auto Scaling group configured for five EC2 instances. Deployments result in performance degradation if capacity falls below four instances. What is the most cost-effective deployment policy to maintain performance?