A company deployed an application on AWS Elastic Beanstalk and set the environment's Auto Scaling group desired capacity to five EC2 instances. If capacity drops below four instances during deployments, application performance degrades. The environment uses the all-at-once deployment policy. What is the MOST cost-effective change to prevent the performance problem during deployments?

The correct option is to change the deployment policy to rolling with additional batch and set the additional batch size to 1. This is the most cost-effective solution because it ensures that during a deployment, at least one additional instance is launched before terminating old instances. This maintains the desired capacity of five instances throughout the deployment, preventing performance degradation without significantly increasing costs. Changing the Auto Scaling group desired capacity to six instances would increase costs permanently, even when not deploying. Changing to traffic splitting is a more complex and potentially more expensive deployment strategy, and an evaluation time of 1 hour is irrelevant to maintaining capacity during the deployment itself. Changing to rolling with a batch size of 2 would still lead to a temporary reduction in available instances if the batch size is applied to the existing instances before new ones are provisioned, potentially dropping below the critical four-instance threshold.