AWS reliability
Reliability is a property of a workload and its dependencies, not a product label. State the business objectives before selecting a topology.
- Availability: define the service-level indicator, measurement window and allowed failure budget.
- RPO: maximum acceptable data loss, measured against a specific failure.
- RTO: maximum acceptable restoration time, including detection, decisions and validation.
Start with backup and recovery drills, architecture review, observability and incident operations.
A read replica, snapshot, Multi-AZ deployment and cross-Region backup address different risks. A snapshot may protect data but not restore DNS, secrets, network rules or compatible application code.
Exercise: enumerate the dependencies of a three-tier application. For each, choose one failure, detection method, recovery owner and validation test. Measure the complete recovery path; do not quote a database failover time as the entire application’s RTO.