⚙️ What is Replication?

Definition:

Replication is the process of creating copies of data across multiple servers/nodes to ensure high availability, durability, and fault tolerance.

• Primary (master) node → handles writes
• Replica (secondary/slave) nodes → serve reads and act as backups

Goal:

• Prevent data loss
• Reduce latency for geographically distributed users
• Improve read scalability

---

🔄 Types of Replication

1. Synchronous Replication

• Writes are applied to primary and replica simultaneously.
• Guarantees strong consistency.
• Trade-off: Higher latency because write waits for all replicas.

2. Asynchronous Replication

• Writes are applied to primary first, then replicas eventually.
• Trade-off: Lower latency, eventual consistency (replicas may lag).

---

🏢 Multi-AZ Replication (Availability Zones)

Availability Zone (AZ) = isolated data center within a region.

• Multi-AZ replication → replicate data across multiple AZs within the same region.
• Purpose: Protect against data center failures.
• Typically synchronous replication to maintain strong consistency.

Example:

Region: us-east-1
AZs: us-east-1a, us-east-1b, us-east-1c

Primary DB → us-east-1a
Replica DBs → us-east-1b, us-east-1c

Advantages:

• Automatic failover if primary AZ goes down
• Low-latency replication within the same region
• High availability SLA (99.99%)

---

🌍 Multi-Region Replication

Region = geographically distant location (different country or continent).

• Multi-region replication → replicate data across regions.

• Purpose:

  • Disaster recovery (DR) in case of regional outage
  • Serve global users with lower latency
• Typically asynchronous replication due to network latency.

Example:

Primary DB → us-east-1 (Virginia)
Replica DB → eu-west-1 (Ireland)
Replica DB → ap-south-1 (Mumbai)

Advantages:

• Disaster recovery across regions
• Reduced read latency for global users
• Compliance with local data regulations

Trade-offs:

• Higher network latency → often eventual consistency
• More complex failover strategies
• Increased cost

---

🔧 Replication Models in Practice

Type	Use Case	Example
Single AZ	Basic HA	MySQL Master-Slave within one AZ
Multi-AZ	High availability & failover	AWS RDS Multi-AZ
Multi-Region Active-Passive	Disaster recovery	Primary in US → Read replica in EU
Multi-Region Active-Active	Low-latency global reads	DynamoDB Global Tables, Cassandra

---

⚡ Key Considerations

1. Consistency

   • Synchronous = strong consistency
   • Asynchronous = eventual consistency

2. Latency

   • Multi-AZ → low (tens of ms)
   • Multi-region → higher (hundreds of ms)

3. Failover

   • Multi-AZ → usually automatic
   • Multi-region → may require manual or orchestrated failover

4. Cost

   • More replicas = higher storage + network cost

5. Use case

   • Multi-AZ → HA within one region
   • Multi-region → global users + disaster recovery

---

🧩 Real-world Examples

System	Replication Strategy
AWS RDS	Multi-AZ synchronous for HA, Multi-region asynchronous for DR
Cassandra	Multi-region active-active, eventual consistency
DynamoDB Global Tables	Multi-region active-active with eventual consistency
MongoDB Atlas	Multi-AZ replica sets and global clusters

---

🧭 Key Takeaways

• Replication = copy data across nodes for durability and availability
• Multi-AZ → same region, synchronous, high availability
• Multi-region → different regions, usually asynchronous, disaster recovery + global latency optimization
• Trade-offs: latency, consistency, cost, complexity

---

💡 Analogy:

• Multi-AZ = “backup in the next building” → fast and reliable
• Multi-region = “backup in another city/country” → slower but protects against disasters

---