Beyond the CAP partition
The CAP theorem describes a choice during a network partition. But there is a quieter tradeoff that applies all the time, even when the network is healthy: the tension between consistency and latency, often summarized by the PACELC framing.
Else, with no partition, a system still chooses between latency and consistency.
Why strong consistency costs latency
To guarantee a read sees the latest write, the system must coordinate across replicas before answering:
- A strongly consistent read may need to reach a quorum or the leader, adding round trips.
- A linearizable system cannot simply answer from the nearest, possibly stale, replica.
Weaker consistency lets a node reply from local state immediately, which is faster but may be stale.
Choosing per operation
Mature systems do not pick one globally. They let each operation choose:
- A bank balance transfer demands strong consistency despite the latency.
- A social feed view tolerates stale data to stay fast.
Key idea
Even without partitions, stronger consistency requires coordination that adds latency, so systems often choose the level per operation.