Reliability and Business Continuity
Recoverability vocabulary
RPO is tolerated data loss; RTO is tolerated downtime. Every design question asks which strategy meets both at the lowest cost. In increasing cost and decreasing RTO the four patterns are backup and restore, pilot light, warm standby, and multi-site active-active.
Data services
RDS Multi-AZ keeps a synchronous standby with automatic failover via DNS; it improves availability, not read throughput - that is what read replicas are for, and they replicate asynchronously, can be promoted, and can be cross-region. S3 provides eleven nines of durability; versioning protects against overwrite and delete, MFA delete hardens it, and cross-region replication requires versioning on both buckets. EBS snapshots are incremental and stored regionally, but each restore is a full volume, so copy snapshots and AMIs for cross-region DR. AWS Backup centralizes schedules, retention, and Vault Lock for compliance.
Routing and scaling
Route 53 failover routing needs a health check on the primary record; without it, no failover occurs. Alias records work at the zone apex and cost nothing for AWS targets. Spread Auto Scaling groups across AZs behind an ELB, and know whether ELB or EC2 health checks drive instance replacement.
Traps
Distractors claim Multi-AZ improves read performance or place a standby in the same AZ. Backup and restore is often correct when RTO is hours and cost matters.
Rule of thumb
Pick the cheapest pattern that satisfies RTO. RTO of one hour with five-minute RPO means warm standby - a scaled-down but running copy with continuous replication - not pilot light, which still needs time to launch instances.