Beyond CAP: PACELC and Modern Perspectives
You'll learn: How the PACELC theorem extends CAP by addressing trade-offs that occur even when the network is healthy.
The Limitation of CAP
The CAP theorem taught us about trade-offs during network partitions (the "P" scenarios). But what about the 99% of the time when your distributed system is running normally with no partitions? CAP is silent on this, yet you're still making critical design choices every single moment.
Enter PACELC
The PACELC theorem fills this gap. It states:
If there is a Partition (P), choose between Availability (A) and Consistency (C); Else (E), when the system is running normally, choose between Latency (L) and Consistency (C).
Think of it like this: CAP handles the crisis (partitions), but PACELC covers everyday operations. Even without network failures, distributed systems must decide: should we wait for all replicas to confirm a write (strong consistency, higher latency) or respond immediately with potentially stale data (lower latency, weaker consistency)?
Why This Matters
Consider a database replicating across three data centers during normal operation:
- High consistency = wait for all replicas to acknowledge → slower responses
- Low latency = respond after one replica → faster, but readers might see old data
Systems like Cassandra are PA/EL (available during partitions, prioritize latency otherwise), while systems like HBase are PC/EC (consistent during partitions, consistent even at the cost of latency during normal operation).
PACELC reveals that the consistency-versus-performance trade-off is constant, not just a partition-time concern.
Key Takeaway:
PACELC extends CAP by acknowledging that distributed systems must also choose between latency and consistency during normal, partition-free operation—not just during failures.