How would you approach multi-region disaster recovery for a security-sensitive system at Amazon Services?
