PA/EC Systems: Mixed Strategies
What You'll Learn
You'll learn how some systems favor availability during network partitions but switch to prioritizing consistency over latency during normal operation.
Understanding PA/EC Systems
PA/EC stands for Partition-Availability / Else-Consistency: systems that choose availability when partitions occur, but prioritize consistency over low latency when the network is healthy.
This might seem contradictory at first — why would an AP system suddenly care about consistency? The answer lies in recognizing that partitions are rare, while normal operation is common. During the 99% of the time when nodes can communicate, these systems deliberately slow down reads or writes to maintain stronger consistency guarantees.
How It Works in Practice
When the network is working:
- Systems wait for multiple nodes to acknowledge writes
- Reads may contact multiple replicas to ensure freshness
- Higher latency is accepted to avoid serving stale data
When partitions happen:
- The system degrades gracefully, accepting writes even if replicas are unreachable
- Availability is preserved, even if it means temporary inconsistencies
- The system relies on eventual consistency and conflict resolution later
Real-World Example
Think of a collaborative document editor like Google Docs. During normal operation, it waits to sync your changes across servers before confirming the save — prioritizing consistency so your teammate sees accurate content. But if your network connection becomes flaky (a partition from your perspective), the editor keeps working, saving locally and accepting that synchronization will happen later. You don't lose availability, but you accept temporary divergence.
Key Takeaway: PA/EC systems are pragmatic — they optimize for the common case (consistency during normal operation) while maintaining availability during rare partition events, accepting higher latency as the price for better consistency when possible.