Mastering High Availability: Multi-AZ and Multi-Region Architectures

This study guide focuses on designing and configuring AWS workloads for maximum resilience. It covers the transition from single-point-of-failure architectures to distributed, global systems that minimize downtime through strategic use of multiple Availability Zones (AZs) and AWS Regions.

---

Learning Objectives

After studying this material, you should be able to:

• Translate business uptime requirements (SLAs) into technical architectures using Multi-AZ and Multi-Region strategies.
• Identify and remediate single points of failure in compute and data layers.
• Configure stateful services (RDS, Aurora, DynamoDB) for synchronous or asynchronous replication across zones and regions.
• Implement traffic routing policies using Route 53 to manage failover events.
• Optimize Recovery Time Objectives (RTO) and Recovery Point Objectives (RPO) based on organizational needs.

---

Key Terms & Glossary

• Availability Zone (AZ): One or more discrete data centers with redundant power, networking, and connectivity in an AWS Region.
• AWS Region: A physical location around the world where AWS clusters data centers.
• RTO (Recovery Time Objective): The maximum acceptable delay between the interruption of service and restoration of service.
• RPO (Recovery Point Objective): The maximum acceptable amount of data loss measured in time (e.g., losing 15 minutes of data).
• Synchronous Replication: Data is written to the primary and secondary locations simultaneously before the write is acknowledged.
• Asynchronous Replication: Data is written to the primary first, then copied to the secondary shortly after.

---

The "Big Idea"

The core philosophy of AWS resilience is "Everything fails, all the time." Instead of trying to build a single indestructible server, we build systems that assume components will fail. By distributing application logic across Availability Zones, we protect against local infrastructure failures (like power outages). By distributing across Regions, we protect against catastrophic geographic events and provide low-latency access to a global user base.

---

Formula / Concept Box

---

Hierarchical Outline

1. High Availability (Multi-AZ)
   • Compute Layer: Use Auto Scaling Groups (ASG) spanning at least two AZs. Use Elastic Load Balancing (ELB) to distribute traffic.
   • Data Layer: Use RDS Multi-AZ for synchronous replication and automatic failover.
   • Benefit: Protects against a single data center failure.
2. Disaster Recovery & Global Reach (Multi-Region)
   • Replication: Use Aurora Global Database or DynamoDB Global Tables for cross-region data sync.
   • Traffic Management: Use Route 53 Health Checks and Failover Routing policies.
   • Storage: Enable S3 Cross-Region Replication (CRR).
3. Stateful vs. Stateless Services
   • Stateless: Easy to scale; sessions should be stored in distributed caches like ElastiCache (Redis).
   • Stateful: Requires careful replication strategy (Synchronous vs Asynchronous).

---

Visual Anchors

Multi-AZ Failover Flow

Latency and Region Mapping

---

Definition-Example Pairs

• Active-Active Failover: A configuration where resources in both locations serve traffic simultaneously.
  • Example: A website using DynamoDB Global Tables where users in London hit eu-west-1 and users in New York hit us-east-1 at the same time.
• Active-Passive Failover: A configuration where one site serves traffic and the other is on standby.
  • Example: An RDS Multi-AZ deployment where the standby instance is only activated if the primary fails.
• Pilot Light: A DR strategy where a minimal version of the environment is always running (usually just the database).
  • Example: Keeping an RDS instance running in a secondary region but keeping the web server AMIs ready to launch only during a disaster.

---

Worked Examples

Example 1: Calculating SLA Downtime

Problem: A business requires "Four Nines" (99.99%) availability. How much downtime is allowed per year? Steps:

1. Total minutes in a year: $$365 \times 24 \times 60 = 525$,600$ minutes.
2. Required Uptime: $525,600 \times 0.9999 = 525,547.44$ minutes.
3. Allowed Downtime: $525,600 - 525,547.44 = 52.56$ minutes per year. Outcome: To achieve this, a Multi-AZ architecture with automated failover is mandatory.

Example 2: Configuring Route 53 Health Checks

Problem: Redirect traffic to a static S3 site if the primary ALB in us-east-1 fails. Steps:

1. Create a Health Check targeting the ALB's DNS name.
2. In Route 53, create a record set with Failover Routing Policy.
3. Set the ALB as the Primary and the S3 Static Website endpoint as the Secondary.
4. Associate the Health Check with the Primary record.

---

Checkpoint Questions

1. What is the main difference between RDS Multi-AZ and RDS Read Replicas regarding data consistency?
2. If your RPO is 0, which replication type MUST you use?
3. True or False: A system can be Highly Available without being Fault Tolerant.
4. Which AWS service would you use to globally route traffic based on the lowest network latency?

▶Click to see answers

1. Multi-AZ uses synchronous replication for high availability; Read Replicas use asynchronous replication for scalability.
2. Synchronous Replication.
3. True. HA ensures the system is up most of the time; FT ensures it stays up even during a component failure with zero impact.
4. Route 53 (Latency Routing Policy).

---

Muddy Points & Cross-Refs

• Multi-AZ vs. Read Replicas: Students often confuse these. Remember: Multi-AZ is for Durability/Availability (Synchronous); Read Replicas are for Scaling (Asynchronous).
• Aurora Replication: Unlike standard RDS, Aurora uses a shared storage volume across AZs, making failover much faster (often < 30 seconds).
• Stateful Failover: Even if the DB fails over, the application might need to reconnect. Always implement Connection Retries in your application code.

---

Comparison Tables

Comparison: Deployment Scopes

[!IMPORTANT] For the DOP-C02 exam, always check if the question asks for minimizing cost or minimizing downtime. Multi-Region is rarely the "cost-effective" answer unless DR is a strict requirement.