Eventual Consistency with Quorums
What you'll learn: How relaxing the quorum condition enables higher availability while accepting temporary stale reads.
The Relaxed Quorum Rule
In the previous lesson, you learned that R + W > N guarantees strong consistency by forcing overlap between read and write sets. But what if you relax this rule?
Allowing R + W ≤ N means read and write quorums don't have to overlap. You might configure N=5, W=2, R=2 (2+2=4, which is ≤5). Now writes succeed faster (only 2 nodes needed), and reads succeed faster too—but there's no guaranteed overlap.
The Trade-off
Without guaranteed overlap, a read quorum might contact replicas that haven't yet received the latest write. You could read stale data temporarily.
Why accept this?
- Higher availability: If 3 nodes are down, you can still read (R=2) and write (W=2) successfully
- Lower latency: Smaller quorums mean faster responses
- Better partition tolerance: The system stays operational during network splits
How It Works in Practice
Imagine N=5 shopping cart replicas with W=1, R=1 (maximum availability):
- User adds item → write goes to 1 replica immediately, returns success
- Other 4 replicas receive updates asynchronously (background replication)
- User refreshes → read might hit a different replica that hasn't caught up yet
- Eventually all replicas converge to the same state
This is eventual consistency: given enough time without new writes, all replicas will agree.
When to Use This
Choose R + W ≤ N when:
- Temporary stale reads are acceptable (shopping carts, session data)
- Availability matters more than instant consistency
- Low latency is critical
Key Takeaway: Setting R + W ≤ N sacrifices strong consistency for higher availability and lower latency, accepting that reads may temporarily return stale data until replicas converge.