A company wants to launch its workload on AWS and requires the system to automatically recover from failure. Which pillar of the AWS Well-Architected Framework includes this requirement?