PACELC Latency-Consistency Tradeoff Lab (Interactive)
See the always-on half of the tradeoff: consistency vs latency when no partition exists. Configure a replication archetype, then sweep region RTT and fsync cost to compute write latency, staleness, and whether you fit the SLO budget.
PACELC Latency-vs-Consistency Dial
CAP only speaks during the 0.1% of time a partition rages. PACELC prices the other 99.9%: Else choose Latency or Consistency.
How It Works Under the Hood
PACELC extends CAP with the observation that Even without partitions, systems trade latency against Consistency: a synchronous cross-region quorum commit is perfectly consistent but pays round-trip light-speed on every write, while async replication answers fast from a local replica that may be seconds behind. This lab models the four archetypes — PC/EC, PA/EL, PC/EL, PA/EC — and computes p99 write latency from fsync plus network terms you control. The gap between the latency bar and the SLO budget is the visual proof that Dynamo-style and Spanner-style stores made opposite always-on choices.
Core Architectural Principles
- Write latency composed from fsync cost plus optional cross-region RTT when quorum-synchronous.
- Staleness metric grows with replication lag the moment you choose EL over EC.
- SLO budget bar shows which PACELC archetype can actually meet your p99 target.
Say CAP only governs partition behavior and PACELC governs the other 99.9% of the time — examiners listen for this. When choosing EL, quantify it: “writes ack in under 5 ms locally, replication SLO 200 ms, staleness p99 under 1 s,” which turns a theorem into a capacity conversation.
Every millisecond you save by not waiting for remote replicas is bought directly from freshness.