In one sentence
Build recoverable systems by removing single points of failure and matching redundancy to RTO and RPO.
Key points
- Use multi-AZ compute, load balancing, Auto Scaling, and appropriate data replication.
- Separate high availability, backup, disaster recovery, and fault tolerance because they solve different failure scopes.
- Choose Multi-AZ for regional availability and Multi-Region when regional failure or global continuity must be addressed.
Exam takeaway
Test restore, failover, dependency recovery, and operational runbooks regularly.
Failure boundaries
- A Region is a geographic failure boundary containing multiple isolated AZs.
- EC2 instances, EBS volumes, and subnets are zonal resources and need cross-AZ redundancy to survive an AZ failure.
- Multi-AZ also requires load balancing, data-tier failover, capacity, and health checks.
- Multi-Region additionally requires replication, routing, RTO / RPO, failover, and exercises.
- Regional managed services may use multiple AZs internally, but resilience must still be verified per service.