Sweaty Imposter

Quorum Reads/Writes

TechniqueCoordinationSeniorStaffUpdated
Write to W replicas, read from R replicas, with W + R > N. Guarantees any read sees at least one of the writes.

When to Use Quorum Reads/Writes

Tunable consistency in a replicated store. Lower W and R = faster + cheaper but eventual; higher = stronger consistency but slower and less available.

In Practice

Cassandra and DynamoDB expose read/write consistency per request. Spanner and CockroachDB use Paxos/Raft (a flavor of quorum) for every write.

How Quorum Reads/Writes Comes Up in Interviews

Scenario

You're running a Cassandra cluster with N=3 replicas. You want any read to be guaranteed to see the most recent successful write, while still tolerating one replica being down. How do you configure R and W?

Quorum reads/writes with W + R > N (here W=2, R=2) ensure read and write sets always overlap on at least one replica, so reads see the latest write. Read-repair is a complementary background mechanism, not the consistency knob itself.

Quorum with W + R > N ensures the read and write sets always share at least one replica, so any read sees the freshest write. With N=3, W=R=2 is the standard quorum that tolerates exactly one replica down. Cassandra exposes this per-query via consistency levels (QUORUM, LOCAL_QUORUM, ONE, etc.), as does DynamoDB. Lowering R or W trades consistency for latency and availability, CL=ONE reads can miss recent writes but survive any two-replica outage. Strong consistency on top of replication usually needs either quorum or a consensus protocol (Paxos, Raft) underneath; read-repair and anti-entropy are convergence mechanisms, not the consistency primitive itself.

Where It Comes Up