The correct option is Rolling with additional batch. This deployment policy temporarily launches extra instances for each batch so the environment remains at full capacity during the rollout and it meets the requirement to avoid any reduction in capacity while keeping additional cost low.

The Rolling with additional batch strategy updates instances in batches and adds a short lived extra batch to preserve the same number of serving instances during each update. This prevents throughput dips during the deployment and keeps extra cost minimal because the additional instances exist only for the duration of the rollout.

Rolling is not appropriate because it updates batches without adding extra instances. That approach reduces available capacity during each batch and can harm user experience under load.

Immutable does provide zero downtime by creating a fresh Auto Scaling group and replacing the fleet, but it duplicates the entire environment and is generally much more expensive. That higher cost makes it a less suitable choice when minimizing extra expense is important.

All at once updates every instance at the same time which causes downtime or service interruption and it therefore violates the requirement to avoid any impact on user experience.

Full AWS Practitioner Certification Question