2.2 Business Continuity

Resiliency Architectures

Multi-AZ vs. Multi-Region

• Multi-AZ: Synchronous replication (usually). Protection against data center failure. High Availability (HA).
• Multi-Region: Asynchronous replication. Protection against regional failure. Disaster Recovery (DR).

Queue and Messaging

• SQS (Simple Queue Service): Decouples components.
  • Dead Letter Queue (DLQ): Handles failed messages after max retries.
• SNS (Simple Notification Service): Pub/Sub. Fan-out to SQS, Lambda, Email.

Replication

Database Replication

• RDS Read Replicas: Async. Used for scaling reads. Can be promoted to Primary.
• Multi-AZ Deployment: Sync. Used for HA. Automatic failover.
• Aurora Global Database: Replicates to other regions with latency < 1s. Fast failover.

Storage Replication

• S3 CRR (Cross-Region Replication): Copy objects to another region. Requires Versioning.
• S3 SRR (Same-Region Replication): Copy objects to another bucket in the same region (e.g., Log Aggregation).

WARNING

Exam Gotcha: S3 Replication is not retroactive. It only replicates objects uploaded after replication is enabled. To replicate existing objects, you must use S3 Batch Replication.