Replication — Sync vs Async

Replication is a core strategy for fault tolerance, scalability, and availability in distributed systems.
It means keeping multiple copies of data across nodes or regions.

1. Why Replication?

High availability → if one node fails, replicas serve traffic.
Read scalability → reads can be distributed across replicas.
Disaster recovery → replicas protect against data center failure.
Low latency → replicas closer to users (geo-replication).

2. Synchronous Replication

How It Works

Write is acknowledged only after all replicas confirm.
Guarantees strong consistency.

Advantages

No stale reads (always up-to-date).
Safe for critical data (e.g., banking).

Disadvantages

Slower writes (must wait for replicas).
If one replica is slow, whole system is slow.
Lower availability (replicas must all be reachable).

Example

Google Spanner (uses TrueTime + synchronous replication).
PostgreSQL synchronous replication.

3. Asynchronous Replication

How It Works

Write is acknowledged after primary node confirms, replicas updated later.
Provides eventual consistency.

Advantages

Faster writes.
Higher availability (replicas can lag).
Good for read-heavy systems.

Disadvantages

Reads may be stale.
Risk of data loss if primary fails before replicas updated.

Example

MySQL async replication.
Cassandra (eventual consistency).

4. Semi-Synchronous Replication

Hybrid approach.
Primary waits for at least one replica before ack.
Balances latency and durability.

5. Trade-offs Summary

Type	Consistency	Latency	Availability	Risk of Data Loss
Sync	Strong	High	Lower	Low
Async	Eventual	Low	High	Higher
Semi-sync	Tunable	Medium	Medium	Medium

6. Interview Tips

Always mention replication for HA & scalability.
Say: “For banking, I’d use synchronous replication. For social media feeds, async is fine.”
Bring up semi-sync as a middle ground.
Tie to CAP theorem: sync → consistency, async → availability.

7. Next Steps

Learn about Failover Strategies.
Explore Geo-replication & Multi-region.