Replication Topologies
What you'll learn: You'll understand how multi-leader replication systems can be organized in different network structures, each with unique trade-offs for propagation speed and complexity.
Core Idea
In multi-leader replication (where multiple nodes accept writes), the topology determines how write operations propagate between leaders. Think of this like choosing how rumor spreads through a friend group: does everyone tell everyone else directly, or does information pass through a chain?
Common Topologies
Circular Topology
Each leader passes writes to exactly one neighbor, forming a ring. Node A sends to B, B to C, C back to A.
Analogy: Like whispering a message around a circle of people—efficient but fragile. If one person leaves, the chain breaks.
- Pros: Simple, predictable paths
- Cons: Single point of failure; high latency (writes must traverse entire ring)
Star Topology
One central "root" leader receives all writes and distributes them to other leaders.
Analogy: Like a news anchor broadcasting to regional stations—centralized control but a critical bottleneck.
- Pros: Simple coordination, easier conflict detection
- Cons: Root node is single point of failure; scaling limited by root's capacity
All-to-All Topology
Every leader sends writes directly to every other leader.
Analogy: Group chat where everyone messages everyone else—fast but chaotic.
- Pros: Lowest propagation latency, fault-tolerant (no single point of failure)
- Cons: Most complex; potential for causality violations if writes arrive out of order
Choosing a Topology
Your choice depends on fault tolerance needs (all-to-all wins), latency requirements (all-to-all for speed), and operational complexity tolerance (circular/star are simpler).
Key Takeaway: Replication topologies determine write propagation paths in multi-leader systems—circular and star are simpler but fragile, while all-to-all offers speed and resilience at the cost of complexity.