Skip to content

Primary SAA curriculum

Multi-Region Disaster Recovery On AWS

Choose and build an AWS disaster recovery strategy from RTO and RPO, progressing from backup and restore through pilot light, warm standby, and active-active.

5 min read

After this, you will understand

This lesson replaces the vague instruction to 'use two Regions' with a decision model that connects business recovery objectives to cost and architecture.

Article guideprerequisites, mental models, and concepts

Article overview

intermediateCloudCertificationReliability

Three useful mental models

In plain terms

Choose the least complex DR strategy that can meet the required recovery time and acceptable data loss, then prove it through failover and failback exercises.

Decision pressure

Teams duplicate compute but forget data and traffic, assume DNS failover is instant, or discover during an outage that the recovery Region lacks capacity, permissions, or a current usable copy.

Exam-ready model

Define the disaster and objectives, select backup and restore, pilot light, warm standby, or active-active, and design data, traffic, capacity, security, testing, and failback as one recovery system.

Think before reading

Why is active-active not automatically the best disaster recovery design?

Its shorter recovery time costs more and introduces harder routing, consistency, deployment, failure-isolation, and operational problems that many workloads do not require.

Connected learning

These lessons add useful context to the current core lesson.
  1. 1Backup vs Replication Recovery DesignAWS Scenario

Multi-Region Disaster Recovery On AWS — Quick Learn

A workload already spans multiple Availability Zones, so it can survive many local infrastructure failures. The remaining requirement is to recover when its AWS Region cannot serve the workload.

Multi-AZ availability and multi-Region disaster recovery solve different failure scopes. A real DR design is not merely “copy data to another Region.” It must restore usable capacity, dependencies, data authority, and traffic within the business targets.

  • RTO sets how long the service may remain unavailable.
  • RPO sets how much recent data may be missing after recovery.

Those targets choose how much of the recovery Region must exist before an incident.

Four Strategies, Increasing Readiness

The four strategies trade a larger always-running recovery footprint for less work and time during an outage.
  • Backup and restore: protected data and infrastructure definitions exist, but the application is not running in the recovery Region. It has the lowest standing cost and the most work during recovery.
  • Pilot light: the critical core—often replicated data and essential services—is current, while much of the application must still be launched or scaled.
  • Warm standby: a complete, functional copy runs at reduced capacity. Recovery scales it, promotes the required data path, validates it, and moves traffic.
  • Active-active: multiple Regions serve production traffic. It can reduce recovery time, but adds the greatest cost and data-consistency, conflict, and routing complexity.

These are design patterns, not universal promises of exact minutes. The chosen RTO/RPO must be tested for the actual workload.

Read deeper about the four DR strategies

What Warm Standby Actually Requires

Data, capacity, dependencies, traffic control, testing, and failback turn a warm standby into a complete recovery system.

In a warm-standby design, Region B is already a functional, observable application environment, just smaller than Region A. Infrastructure as code keeps the environments reproducible, but it does not replicate live application data.

Data movement remains service-specific:

  • Aurora Global Database can replicate relational data across Regions.
  • DynamoDB global tables replicate table items between Regions.
  • S3 Cross-Region Replication copies eligible objects.
  • Cross-Region backup copies preserve historical recovery points.

Most cross-Region replication is asynchronous. The lag at the incident determines the potential data gap. Replication also copies valid-looking bad changes, so backups remain necessary for historical recovery.

A recovery Region needs more than compute and data. It may need regional KMS keys and grants, secrets, certificates, queues, topics, buckets, alarms, service quotas, identity access, network paths, and third-party allowlists. If those exist only in the failed Region, the “standby” is not independently usable.

Read deeper about regional dependency completeness

Failover Is A Controlled Authority Transfer

A safe warm-standby failover is a sequence, not a DNS click:

Detect and decide
↓
Fence unsafe writes in Region A
↓
Inspect replication lag and promote Region B's data path
↓
Scale Region B and validate critical operations
↓
Shift user traffic
↓
Region B becomes authoritative

Route 53 failover routing changes DNS answers based on policy and health. DNS TTLs, resolvers, and existing connections mean traffic does not move instantly.

Amazon Application Recovery Controller (ARC) routing controls provide an explicit, highly available switch with safety rules for moving traffic. ARC does not replace health monitoring or the recovery runbook.

Fencing prevents split brain, where both Regions accept conflicting writes. Exactly one Region should hold write authority unless the application was deliberately designed for multi-writer conflict handling.

Failback Completes The Design

When Region A becomes healthy again, it may no longer contain the newest data because Region B accepted writes during recovery. Do not simply reverse DNS.

Rebuild or resynchronize Region A from the authoritative Region B, validate it, choose a controlled return window, fence writes as required, and then move traffic. Design and test failback before the outage; improvising it after failover is where data authority becomes ambiguous.

Read deeper about the failover and failback lifecycle

SAA Recognition And Traps

  • Region-level recovery with low standing cost and tolerant RTO → backup and restore.
  • Keep only the critical core ready → pilot light; substantial launch and scale work remains.
  • A complete reduced-capacity environment must already run → warm standby.
  • Both Regions actively serve production → active-active, accepting extra consistency and operational complexity.
  • Move users after a regional failure → Route 53 can change DNS answers; ARC can provide controlled routing decisions and safety checks.
  • Very low RPO across Regions → inspect the chosen service's replication behavior and actual lag; “multi-Region” alone is not an RPO guarantee.

Do not assume Multi-AZ survives a Region failure, infrastructure as code copies data, replication replaces backups, DNS changes are instantaneous, or failback is just failover in reverse.

One-Minute Review

Business impact → RTO and RPO
↓
Choose backup/restore, pilot light, warm standby, or active-active
↓
Prepare application capacity + service-specific data + regional dependencies
↓
Test the recovery path
↓
Fence old writes, promote and validate recovery, then move traffic
↓
Rebuild from the new authority before failback

If you remember only one thing: multi-Region DR is a tested transfer of service and data authority, not merely a second copy of the data.

Finished reading?

Your reading history is saved in this browser so you can continue later.

Recommended Next

On-Premises Migration To AWSAWS Architecture Scenarios16 min read

This applies the foundation mental models to a real architecture decision instead of a service inventory.

Optional exploration

These links add context, but they do not replace the recommended next lesson.

Arcflow Plus is coming — review drills, research breakdowns, more AI. Get one email at launch.