The idea
Leaderless systems replicate each value to N nodes. Instead of a leader deciding, a write is sent to all N and the client waits for W confirmations. A read queries all N and waits for R responses.
The magic is the quorum condition: if W plus R is greater than N, then any read set and any write set must overlap on at least one node. That overlapping node holds the latest write, so the reader can find it.
Picking the numbers
- A common choice is N equals 3 with W and R both equal to 2, tolerating one node being down.
- Larger W means stronger durability but slower writes. Larger R means fresher reads but slower reads.
- The reader uses version numbers to pick the newest value among the R responses.
What it does not fix
Quorums tolerate node failures gracefully but still face concurrent writes that need conflict resolution, and edge cases where reads can briefly see stale data despite the overlap rule.
Key idea
A quorum guarantees fresh reads without a leader by forcing read and write sets to overlap whenever W plus R exceeds N.