The 'Else' Clause: Normal Operation Tradeoffs
What you'll learn: Even when no network partition exists, distributed systems must still trade off between consistency and latency.
The Missing Half of the Story
CAP Theorem focuses on what happens during a partition—when you must choose between consistency and availability. But what about the other 99% of the time when your network is functioning normally?
This is the "else" clause: When partitions aren't happening, you still face a fundamental tradeoff between consistency and latency.
The Normal Operation Tradeoff
Imagine you're running a distributed database with replicas in New York and London. A user in Boston writes data. To maintain strong consistency, you must:
- Send the write to all replicas
- Wait for acknowledgments from each one
- Only then confirm success to the user
That round-trip to London adds ~80ms of latency—every single time, even though your network is working perfectly.
The alternative? Return success after writing locally, then asynchronously replicate. You get low latency but risk serving stale data from London until replication completes. This is your consistency-latency tradeoff in normal operation.
Why This Matters
PACELC extends CAP by saying: "If Partition, choose Availability or Consistency; Else (no partition), choose Latency or Consistency."
Systems like Cassandra can be configured for low latency (read/write local replicas) or strong consistency (coordinate across all replicas)—even when everything is working fine. This isn't about failure handling; it's about everyday performance characteristics.
Key Takeaway: The 'Else' clause reveals that consistency vs latency tradeoffs dominate your system's behavior during normal operation, not just during rare partition events—making them arguably more important to your day-to-day design decisions than CAP alone suggests.