How CP and AP Systems Work

Core Concept
CP systems use leader-based replication with quorum protocols. AP systems use leaderless replication with tunable consistency. The mechanisms differ fundamentally.
HOW CP SYSTEMS WORK
A single leader accepts all writes. The leader replicates each write to followers and only commits when a majority acknowledges. In a 5 node cluster, majority is 3. Writes require 2 follower acknowledgments before returning success.
During a partition isolating 2 nodes from 3, only the 3 node side can form a majority. The isolated 2 nodes refuse writes and return errors. Clients see reduced availability, but the system never creates conflicting data.
Latency cost: Inter-region RTT of 50ms means writes take 50ms to 100ms. Within a region (1-2ms RTT), writes cost 5ms to 10ms at p99.
HOW AP SYSTEMS WORK
Data replicates to N nodes (commonly 3). You configure write consistency W and read consistency R. Setting W=1 and R=1 gives maximum availability: writes return after one node acknowledges (1-2ms), reads return from whichever replica responds first.
During partitions, all nodes continue accepting requests locally. Clients get sub-10ms responses even when zones cannot communicate. The trade-off: writes in an isolated zone will not immediately appear elsewhere. Systems reconcile later using background sync processes that compare replica states and fix inconsistencies.
Key Insight: CAP trade-offs apply during partitions. When the network is healthy, many systems provide both consistency AND availability. The choice only matters when failures occur.
THE NUMBERS
CP write latency: 5-30ms within region (quorum). AP write latency: 1-5ms (single node). This latency difference is measurable in every request.

💡 Key Takeaways

✓CP systems use majority quorum: in 5 node cluster, 3 must agree. Isolated minority (2 nodes) refuses writes to prevent conflicts.

✓CP write latency within region is 5ms to 10ms at p99 for consensus. Cross region adds 50ms to 150ms RTT overhead.

✓AP systems with W equals 1, R equals 1 provide 1ms to 2ms write latency and continue working even when most replicas are unreachable.

✓AP systems reconcile conflicts after partitions heal using anti entropy, hinted handoff, and version vectors or last write wins.

📌 Interview Tips

1When explaining CP vs AP, use concrete latency numbers: "CP adds 5-30ms for quorum writes, AP returns in 1-5ms but may serve stale data."

2If asked how leader election works, explain: "Nodes vote, majority wins, each election increments a term number. Nodes reject commands from stale leaders."

3For AP systems, mention conflict resolution: "Last-write-wins risks data loss with clock skew. Version vectors detect conflicts for application-level merging."

← Back to CAP Theorem Overview