Domain 1 of 4 · Chapter 3 of 5

Reliable and resilient multi-account architectures

Unlock the complete study guide + 1,040 practice questions across 16 full exams.

Bundled into the existing AWS Certified Solutions Architect - Professional premium course — no separate purchase.

Included in this chapter:

  • The four DR strategies: RTO/RPO, cost, and effort in depth
  • Static stability and cell-based (bulkhead) fault isolation
  • Application Recovery Controller: routing controls, readiness, zonal shift
  • Cross-Region data replication: Aurora Global, DynamoDB, S3 CRR
  • Org-wide AWS Backup and service quotas as a reliability risk

AWS disaster recovery strategies by RTO, RPO, and cost

DimensionBackup & restorePilot lightWarm standbyMulti-site active/active
Typical RTOHours (redeploy + restore)Tens of minutes (turn on + scale)Minutes (scale up only)Near zero (already serving)
Typical RPOBackup interval (hours)Near zero (continuous replication)Near zero (continuous replication)Near zero (continuous replication)
DR-Region stateData backups only; no running infraData live; app servers switched offScaled-down copy always runningFull capacity, serving traffic
Relative costLowestLowMediumHighest
Serves traffic immediately?No (must redeploy)No (must turn on + scale)Yes, at reduced capacityYes, at full capacity
Traffic / failover mechanismRoute 53 / ARC after redeployRoute 53 DNS failover; ARC routing controlRoute 53 DNS failover; ARC routing controlRoute 53 latency/geoproximity; Global Accelerator

Decision tree

Hours of downtime (RTO)acceptable?Yes (cheapest wins)NoBackup & restoreAWS Backup cross-Region copy +IaC redeploy; RPO = backup intervalNeed near-zero RTO,serve from every Region?Yes (budget high)NoMulti-site active/activefull capacity both Regions;Route 53 latency / GlobalAccelerator; highest costOK to run a scaled-downcopy 24/7 (serve at once)?No (no idle compute)YesPilot lightdata replicated live; app off.Turn on + scale before serving;near-zero RPOWarm standbyscaled-down, always running.Serves at reduced capacity now,scale up only; near-zero RPOAlways: fail over with data-plane controls (ARC routing controls + safety rules),pre-provision capacity (static stability).Raise DR-Region service quotas ahead of time; replication is not backup (keep versioned / point-in-time copies).
Cheat sheet

Cheat sheet

  • Four DR strategies on one RTO/cost spectrum
  • Pilot light keeps compute off; warm standby keeps it scaled-down and running
  • Backup & restore is the only strategy with an hours-scale RPO
  • Choose active/active only when you genuinely serve both Regions
  • Static stability means surviving an event without making changes
  • Prefer data-plane toggles over control-plane edits for failover
  • Cells (bulkheads) shrink blast radius below the Region
  • Match the failure to the right nested isolation boundary
  • ARC routing controls are data-plane on/off switches for Region failover
  • ARC safety rules block invalid failover states
  • ARC readiness checks are a steady-state prep tool, not in the failover path
  • Zonal shift is manual, lasting from 1 minute to 72 hours
  • Zonal autoshift is AWS-initiated and requires weekly practice runs
  • Pre-scale the surviving AZs before any zonal shift
  • Aurora Global Database: sub-second replication, up to 10 secondary Regions
  • Aurora switchover (planned, no data loss) vs failover (unplanned outage)
  • DynamoDB global tables: MREC default vs MRSC for RPO 0
  • S3 CRR needs versioning; RTC's SLA is 99.9% within 15 minutes
  • Replication copies corruption too, so it is not a backup
  • AWS Backup org backup policies push plans down the OU tree
  • AWS Backup cross-account copy protects the last copy from a rogue account
  • Cross-account backup copy needs a customer-managed KMS key
  • DR-Region service quotas are a hidden reliability single point of failure
  • Set the ASG health-check grace period above app startup time
  • RDS Multi-AZ DB cluster adds two readable standbys and sub-35s failover
  • Vault Lock compliance mode makes backups immutable to everyone, even AWS
  • Restore Intelligent-Tiering archive objects before they can replicate

Unlock with Premium — includes all practice exams and the complete study guide.

Also tested in

References

  1. Disaster recovery options in the cloud Whitepaper
  2. Use bulkhead architectures to limit scope of impact Well-Architected
  3. AWS service quotas
  4. What is Amazon Application Recovery Controller (ARC)?
  5. What is AWS Backup?
  6. Backup policies (AWS Organizations)
  7. Creating backup copies across AWS accounts
  8. Zonal autoshift in ARC
  9. Zonal shift in ARC
  10. Using Amazon Aurora Global Database
  11. Using switchover or failover in Amazon Aurora Global Database
  12. Global tables - multi-Region replication for DynamoDB
  13. Replicating objects with S3 Replication
  14. Meeting compliance requirements with S3 Replication Time Control