Quorum Reads & Writes
In distributed databases, quorum reads/writes ensure consistency across replicas.
Instead of requiring all nodes to agree, a quorum allows progress with a majority of nodes.
1. What is a Quorum?
A quorum is the minimum number of votes needed to make a decision in a distributed system.
For reads/writes:
- Read Quorum (R) = number of replicas that must respond for a read.
- Write Quorum (W) = number of replicas that must acknowledge a write.
Given N replicas:
- For strong consistency →
R + W > N. - Ensures at least one replica in read quorum overlaps with write quorum.
2. Example
Let’s say N = 3 replicas.
Strong Consistency:
W = 2,R = 2.- At least one replica is guaranteed to have latest write.
High Availability (weaker consistency):
W = 1,R = 1.- Faster, but reads may return stale data.
3. Trade-offs
| Setting | Pros | Cons |
|---|---|---|
| Strong (R+W > N) | Consistent reads guaranteed | Higher latency, lower availability |
| Weak (R+W ≤ N) | Fast, highly available | Reads may be stale |
4. Real-World Databases
- Cassandra / DynamoDB → tunable consistency: choose R & W per query.
- MongoDB → read preferences (primary, secondary, majority).
- CockroachDB / Spanner → quorum-based with strong consistency.
5. Interview Tips
- Mention quorum when asked about consistency in replication.
- Say: “If I want strong consistency, I’d ensure R + W > N. If I want lower latency, I’d relax this.”
- Bring up tunable consistency (Cassandra) as a real-world example.
- Always tie back to CAP theorem trade-offs.
6. Diagram (N=3, Quorum Example)
Replicas: [ R1 ] [ R2 ] [ R3 ]
Write (W=2): R1 + R2 must ack.
Read (R=2): R2 + R3 must respond.
Overlap → ensures latest data read.7. Next Steps
- Learn about Leader Election & Failover.
- Explore Eventual vs Strong Consistency.