PACELC Framework Explained
What you'll learn: How PACELC extends CAP by considering trade-offs during both partition and normal operation scenarios.
The Complete Formula
CAP Theorem only describes what happens during a network partition: choose Availability or Consistency. But what about the 99.9% of the time when everything works fine?
PACELC gives you the full picture:
"If there's a Partition, choose between Availability and Consistency; Else (during normal operation), choose between Latency and Consistency."
Why This Matters
During normal operation (no partition), distributed systems still face a fundamental trade-off. To keep data consistent across replicas, you must wait for synchronization—which increases latency. To minimize latency, you can read from local replicas without synchronizing—which risks serving stale data (lower consistency).
The Two-Part Decision
-
PA/PC (during partition):
- PA systems (like Cassandra): Stay available, accept inconsistency
- PC systems (like banking): Stay consistent, reject requests
-
EL/EC (normal operation):
- EL systems: Optimize for speed, eventual consistency
- EC systems: Synchronize writes, stronger guarantees
Example: Full Classifications
- PA/EL: Cassandra, DynamoDB (available during partition, low-latency during normal ops)
- PC/EC: Traditional RDBMS with sync replication (consistent always, higher latency)
- PA/EC: Some DNS systems (available during partition, but consistent when stable)
Key Takeaway: PACELC recognizes that distributed systems make trade-offs all the time, not just during rare partition events—latency vs consistency is the everyday battle.