10.1 Disaster Recovery, Cross-Region Replication & Backup Strategies

10.1 Disaster Recovery, Cross-Region Replication & Backup Strategies

Designing resilient data platforms on AWS requires a rigorous understanding of disaster recovery (DR) objectives, automated backup capabilities, and cross-Region data replication patterns. In enterprise data engineering, a regional service disruption or catastrophic failure must not lead to unrecoverable data loss or prolonged platform downtime. AWS provides both centralized management services and native database/storage replication mechanisms to satisfy strict business continuity requirements.

---

1. Core Disaster Recovery Metrics: RPO and RTO

Every disaster recovery strategy is defined by two fundamental business requirements:

• Recovery Point Objective (RPO): The maximum acceptable amount of data loss measured in time. An RPO of 5 minutes means the platform must be able to recover data up to 5 minutes prior to the failure event.
• Recovery Time Objective (RTO): The maximum acceptable duration of infrastructure downtime before business operations are restored. An RTO of 1 hour means service availability must be recovered within 60 minutes of disaster declaration.

The Four Standard AWS DR Architecture Patterns

AWS categorizes DR strategies into four primary architectural patterns, offering trade-offs between cost, operational complexity, and RPO/RTO objectives. The values below are planning illustrations, not AWS service guarantees; derive tested objectives from the workload and chosen services:

---

2. Centralized Data Protection with AWS Backup

AWS Backup is a fully managed, policy-based backup service that centralizes and automates data protection across AWS services. Rather than writing custom scripts for EBS snapshots, RDS backups, or S3 replication, engineers configure unified backup policies called Backup Plans.

Core Capabilities for Data Engineers

1. Supported Native Services: S3, DynamoDB, RDS, Aurora, Redshift, EBS, EFS, AWS Storage Gateway, and other services listed in the current AWS Backup feature matrix. Verify service and feature support by Region before designing a plan.
2. Cross-Account and Cross-Region Copies: Backup Plans can automatically copy backup snapshots to a secondary AWS Region and a separate, isolated target AWS Account (managed within AWS Organizations). This protects backups against compromised primary account credentials or regional disasters.
3. AWS Backup Vault Lock: Implements Write-Once-Read-Many (WORM) immutability on backup vaults. Once Vault Lock is enabled in compliance mode, even the AWS account root user cannot delete or alter backups before the retention period expires, meeting strict SEC 17a-4 and FINRA compliance rules.
4. Transition to Cold Storage: Backups for supported services (such as EFS or DynamoDB) can automatically transition from warm storage to cost-effective cold storage after a configurable number of days.

{
  "BackupPlanName": "EnterpriseDataPlatformDRPlan",
  "Rules": [
    {
      "RuleName": "DailyCrossRegionBackup",
      "TargetBackupVault": "PrimaryVault",
      "ScheduleExpression": "cron(0 1 * * ? *)",
      "StartWindowMinutes": 60,
      "CompletionWindowMinutes": 180,
      "Lifecycle": {
        "MoveToColdStorageAfterDays": 30,
        "DeleteAfterDays": 365
      },
      "CopyActions": [
        {
          "DestinationBackupVaultArn": "arn:aws:backup:us-west-2:111122223333:backup-vault:DRVault",
          "Lifecycle": {
            "DeleteAfterDays": 365
          }
        }
      ]
    }
  ]
}

---

3. Amazon S3 Cross-Region Replication (CRR)

For object data stored in Amazon S3, Cross-Region Replication (CRR) provides automatic, asynchronous copying of objects across S3 buckets in different AWS Regions.

Prerequisites and Technical Requirements

• Versioning: S3 Versioning must be enabled on both the source and destination buckets.
• IAM Permissions: The S3 service must be granted an IAM role allowing s3:GetObjectVersionForReplication on the source and s3:ReplicateObject on the destination.
• KMS Key Ownership: If objects are encrypted with SSE-KMS, the replication configuration must explicitly enable KMS object replication and supply destination KMS key ARNs.

Advanced CRR Features for Analytics

• S3 Replication Time Control (S3 RTC): An enterprise SLA feature that guarantees 99.9% of new objects are replicated to the target Region within 15 minutes. It includes detailed Amazon CloudWatch metrics (ReplicationLatency, PendingObjects, PendingBytes) for compliance tracking.
• S3 Batch Replication: Allows data engineers to replicate existing objects created prior to the configuration of a CRR rule, or objects that previously failed replication.
• Delete Marker Replication: Can be disabled to prevent accidental or malicious object deletion in the source bucket from propagating to the DR bucket.

---

4. Database Cross-Region Replication Mechanisms

Analytical and transactional databases require specialized replication architectures to enforce tight RPO boundaries during regional failures.

Amazon Aurora Global Databases

For high-performance OLTP applications, Aurora Global Databases decouple database compute from storage replication:

• Dedicated storage-level replication infrastructure handles cross-Region data transport with zero impact on primary compute performance.
• Typically delivers cross-Region replication latency under 1 second, but RPO during an unplanned outage depends on observed replication lag.
• Supports up to 10 secondary Regions, subject to engine-version and Region support; per-cluster replica quotas must be checked for the selected engine.
• Provides planned switchover for coordinated role changes and unplanned failover for outages; recovery time and possible data loss depend on topology, lag, and the failure.

# AWS CLI command to execute managed failover for an Aurora Global Database
aws rds failover-global-cluster \
    --global-cluster-identifier enterprise-aurora-global-db \
    --target-db-cluster-identifier arn:aws:rds:us-west-2:111122223333:cluster:aurora-dr-secondary-cluster

Amazon DynamoDB Global Tables

DynamoDB Global Tables provide a fully managed, multi-Region, active-active database replication framework:

• Utilizes DynamoDB Streams under the hood to propagate write operations across replica tables in target Regions.
• Employs Last-Writer-Wins (LWW) timestamp resolution to reconcile concurrent writes occurring in different Regions.
• Applications can read and write locally in any participating Region, offering near-zero RTO and minimal latency for globally distributed data pipelines.

Amazon Redshift Cross-Region Snapshot Copy

For analytical data warehouses, Amazon Redshift supports Automated Cross-Region Snapshot Copy:

• Redshift automatically copies cluster snapshots to a secondary Region as soon as they are taken.
• KMS master keys in the destination Region are mapped to encrypt the snapshot data in the target location.
• In a DR event, a replacement Redshift cluster or Redshift Serverless namespace is restored directly from the copied snapshot.