What's changed: Service coverage: added catalog sections (s4 compute/DB/storage/network, s5 analytics/ML/IoT/media/integration)
2.1High Availability and Disaster Recovery
Understand designs for availability and resilience—Multi-AZ/Multi-Region, RTO/RPO, the four DR strategies (backup & restore/pilot light/warm standby/multi-site), and Route 53 failover.
Availability starts with AZ redundancy; for disasters you pick among four strategies based on RTO/RPO. Cost and recovery speed trade off.
2.1.1Choosing a DR strategy
- RTO/RPO: RTO = tolerable time to recover, RPO = tolerable data loss (time); requirements drive the strategy.
- Backup & restore: cheapest; RTO/RPO in hours.
- Pilot light: only the core (e.g., replicated DB) runs minimally; scale out on disaster; RTO in tens of minutes.
- Warm standby: a scaled-down copy always running; scale up fast; RTO in minutes.
- Multi-site (active-active): full capacity in multiple regions; near-zero RTO/RPO; most expensive.
Most common on SAP-C02: cheapest/slow = backup & restore, core-only = pilot light, scaled-down always-on = warm standby, full multi-region = multi-site, plus RTO = time / RPO = data. Shorter RTO/RPO requirements push you to higher (costlier) strategies.
Route 53 health checks + failover routing auto-switch to standby on primary failure. Replicate data with RDS cross-region read replicas or DynamoDB global tables.
Establish availability first within one region via Multi-AZ (ELB + Auto Scaling across AZs, RDS Multi-AZ, 3-AZ-redundant S3/DynamoDB), then choose multi-region DR by requirement for regional failures. RTO = tolerable time to recover; RPO = tolerable data loss (time). The four strategies trade cost vs speed: backup & restore (AWS Backup, cross-region snapshot copy; RTO/RPO in hours) < pilot light (DB continuously replicated via cross-region read replicas, app stopped/minimal and started on disaster; tens of minutes) < warm standby (a scaled-down full stack always on, scale up fast; minutes) < multi-site/active-active (production in multiple regions, DynamoDB global tables; near-zero). Switch with Route 53 health checks + failover/latency/weighted routing, deliver via CloudFront, and reproduce the standby with IaC (CloudFormation/StackSets). For RPO≈0 use active-active or synchronous replication; for cost, a lower strategy—derive it from the RTO/RPO numbers.
| Strategy | Standby | RTO/RPO | Cost |
|---|---|---|---|
| Backup & restore | None (backups only) | Hours | Lowest |
| Pilot light | Core (DB) only | Tens of min | Low |
| Warm standby | Scaled-down, always on | Minutes | Medium |
| Multi-site | Full, multi-region | Near-zero | Highest |
Scenario: meet RPO of 1 min and RTO of 5 min for payments while limiting cost. → Choose warm standby: keep a scaled-down full stack always on in another region, continuously replicate the DB (cross-region replica or global tables), scale to production capacity with Auto Scaling on disaster, and switch via Route 53 failover. For stricter RPO≈0, move up to active-active.
FAQ: Q. RTO vs RPO? → A. RTO is how fast you recover (time); RPO is how much data loss you tolerate (time span). Q. Pilot light vs warm standby? → A. Pilot light keeps only the core (mainly DB) on and starts the app on disaster; warm standby keeps a scaled-down full stack always on for fast scale-up.
Trap: “choose multi-site (active-active) when cost is the top priority” is wrong—cheapest is backup & restore (RTO/RPO in hours). Conversely “meet RPO≈0 with backup & restore” is wrong—near-zero needs synchronous replication/active-active. Derive the strategy from the RTO/RPO numbers.
2.1.2Section summary
- DR = backup&restore < pilot light < warm standby < multi-site (cost/speed)
- RTO = recovery time / RPO = data loss; failover = Route 53
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. For a mission-critical system with near-zero RPO and RTO of minutes, which DR strategy fits?
Q2. Which DR strategy prioritizes cost and tolerates RTO/RPO of several hours?
Q3. On primary-region failure, you want to auto-direct user traffic to a standby region. What?