What does W + R > N actually guarantee?

That the set of replicas a write lands on and the set a read consults must overlap in at least one node — so a read is guaranteed to touch a replica that already has the latest acknowledged write and can return it. N is the replica count, W the number of replicas that must ack a write, R the number that must answer a read. If W + R ≤ N the two sets can be disjoint and a read can miss the newest write. The Dynamo paper states it directly: "Setting R and W such that R + W > N yields a quorum-like system."

Is partition tolerance optional in the CAP theorem?

No. A network partition is a fact you cannot design away — packets get dropped or delayed — so a real distributed system must tolerate P. That means the actual choice CAP forces is between consistency and availability, and only during a partition. Eric Brewer: "The '2 of 3' formulation was always misleading… CAP prohibits only a tiny part of the design space: perfect availability and consistency in the presence of partitions, which are rare."

What is the difference between CAP and PACELC?

CAP only describes the partition case. PACELC adds the far more common no-partition case: even when the network is healthy, a replicated system trades consistency against latency. Daniel Abadi: "if there is a partition (P), how does the system trade off availability and consistency (A and C); else (E), when the system is running normally in the absence of partitions, how does the system trade off latency (L) and consistency (C)?"

What is a sloppy quorum?

A relaxation of the strict W-of-N rule so a system stays writable during failures. Instead of requiring the specific replicas that own a key, a sloppy quorum accepts the write on the first N healthy nodes it can reach and hands it off later. Dynamo: "it does not enforce strict quorum membership and instead it uses a 'sloppy quorum'; all read and write operations are performed on the first N healthy nodes from the preference list." It buys availability at the cost of the strict overlap guarantee — the write may not land where a later read looks.

Which consistency model does etcd, ZooKeeper, or Cassandra use?

etcd and ZooKeeper are CP systems built on a majority-quorum consensus log: etcd "ensures linearizability for all other operations by default"; ZooKeeper gives sequential consistency for writes but reads can be stale unless you call sync(). Cassandra is tunable — it "supports a per-operation tradeoff between consistency and availability through Consistency Levels," so reading and writing at QUORUM gives W+R>RF and strong reads, while ONE gives eventual consistency and lower latency.

Consistency, Quorums & CAP

Replication forces copies to disagree. The consistency ladder, the N/W/R quorum rule (W+R>N), and CAP stated precisely — the real choice is C-vs-A only during a partition — turned live on a 5-replica cluster.

Concept · Systems. The source ↗

A free, interactive, animated visual explainer of Consistency, Quorums & CAP — built to be understood, not skimmed.

Questions

What does W + R > N actually guarantee?: That the set of replicas a write lands on and the set a read consults must overlap in at least one node — so a read is guaranteed to touch a replica that already has the latest acknowledged write and can return it. N is the replica count, W the number of replicas that must ack a write, R the number that must answer a read. If W + R ≤ N the two sets can be disjoint and a read can miss the newest write. The Dynamo paper states it directly: "Setting R and W such that R + W > N yields a quorum-like system."
Is partition tolerance optional in the CAP theorem?: No. A network partition is a fact you cannot design away — packets get dropped or delayed — so a real distributed system must tolerate P. That means the actual choice CAP forces is between consistency and availability, and only during a partition. Eric Brewer: "The '2 of 3' formulation was always misleading… CAP prohibits only a tiny part of the design space: perfect availability and consistency in the presence of partitions, which are rare."
What is the difference between CAP and PACELC?: CAP only describes the partition case. PACELC adds the far more common no-partition case: even when the network is healthy, a replicated system trades consistency against latency. Daniel Abadi: "if there is a partition (P), how does the system trade off availability and consistency (A and C); else (E), when the system is running normally in the absence of partitions, how does the system trade off latency (L) and consistency (C)?"
What is a sloppy quorum?: A relaxation of the strict W-of-N rule so a system stays writable during failures. Instead of requiring the specific replicas that own a key, a sloppy quorum accepts the write on the first N healthy nodes it can reach and hands it off later. Dynamo: "it does not enforce strict quorum membership and instead it uses a 'sloppy quorum'; all read and write operations are performed on the first N healthy nodes from the preference list." It buys availability at the cost of the strict overlap guarantee — the write may not land where a later read looks.
Which consistency model does etcd, ZooKeeper, or Cassandra use?: etcd and ZooKeeper are CP systems built on a majority-quorum consensus log: etcd "ensures linearizability for all other operations by default"; ZooKeeper gives sequential consistency for writes but reads can be stale unless you call sync(). Cassandra is tunable — it "supports a per-operation tradeoff between consistency and availability through Consistency Levels," so reading and writing at QUORUM gives W+R>RF and strong reads, while ONE gives eventual consistency and lower latency.

Consistency, Quorums & CAP

Questions

Related explainers