Name: System Design Masterclass
Availability: InStock

Over the last 30 lessons, you've learned about data partitioning, modes, topologies, CDC, geographic distribution, and failover configurations. Each lesson covered a specific technique. But in a real system design interview or architecture review, you need to combine them.

A senior engineer doesn't just know what is. They know WHEN to use it (distributing data across nodes), WHEN to combine it with replication (storing copies on multiple ring nodes), and WHEN the simpler alternative is better (just use range partitioning if you need range queries).

This lesson is the capstone for the module. We'll build a decision framework that maps your system's requirements to the right combination of replication techniques.

Start with three questions:

1. What's your read-to-write ratio?

Ratio	Strategy
Read-heavy (>10:1)	Star topology with read replicas. Most web apps live here.
Balanced (~1:1)	Consider for write scaling. Read replicas may not help.
Write-heavy (<1:10)	Shard aggressively. Replicas for DR only, not read scaling.

2. How many regions do your users span?

Geography	Strategy
Single region	Multi-AZ active-passive. Simple, reliable, sufficient.
2 regions	Active-passive with read replica in second region. Or active-active if writes are -sensitive in both.
3+ regions	Multi-region deployment with geo-partitioned writes. Active-active with conflict avoidance.

3. What's your tolerance for data loss?

RPO Requirement	Mode
Zero data loss	Synchronous replication (same region only, cross-region sync is too slow)
Seconds acceptable	Semi-synchronous or async with close replicas
Minutes acceptable	Async with cross-region replicas. Log shipping for DR.

Combine the answers and you get a topology:

Replication Topology

The rest of this lesson

Continue this lesson

Pulling It All Together

The Decision Framework