3.1.1.4. Techniques to Achieve High Availability (Multi-AZ, Multi-Region)

3.1.1.4. Techniques to Achieve High Availability (Multi-AZ, Multi-Region)

High availability is an architecture decision, not a checkbox. Each technique adds cost and complexity — the goal is matching the HA level to the business requirement.

Multi-AZ (standard for most production workloads):

• Deploy compute across 2+ AZs via ASG or ECS
• Use ALB/NLB to distribute traffic across AZs
• Database: RDS Multi-AZ or Aurora (automatic)
• Cost: Minimal additional (cross-AZ data transfer is small)
• Protects against: AZ failure, hardware failure, maintenance events

Multi-Region (for mission-critical or global workloads):

• Duplicate infrastructure in 2+ regions
• Route 53 health checks route traffic away from unhealthy regions
• Data replication: Aurora Global Database, DynamoDB Global Tables, S3 CRR
• Cost: Significant (full duplicate infrastructure + cross-region data transfer)
• Protects against: Regional outage, natural disaster, compliance (data residency)

Active-Passive vs. Active-Active:

• Active-Passive: Secondary region on standby. Lower cost, higher RTO (must promote/scale secondary).
• Active-Active: Both regions serve traffic. Higher cost, near-zero RTO. Requires data conflict resolution (DynamoDB Global Tables handles this with last-writer-wins).

Exam Trap: Multi-AZ load balancers (ALB/NLB) only distribute within a single region. For multi-region traffic distribution, you need Route 53 (DNS-level routing) or Global Accelerator (network-level routing). If a question asks about failover between regions, ALB is not the answer.