How does Cassandra's ring architecture and consistent hashing work?

Expected Interview Answer

Step-by-Step Explanation

1. Step 1

   Hash the partition key

   Cassandra applies the partitioner (Murmur3 by default) to the partition key, producing a 64-bit token.

2. Step 2

   Map token to ring position

   The token is a point on a circular token space that spans from -2^63 to 2^63-1.

3. Step 3

   Find the owning node

   Walking the ring clockwise, the first node whose token range covers that value owns the primary replica.

4. Step 4

   Use virtual nodes

   Each physical node holds many small token ranges (vnodes), so ownership is spread in tiny slices for balance.

5. Step 5

   Rebalance on topology change

   Adding or removing a node only reassigns the token ranges near it, moving a fraction of the data.

Code Example

CREATE TABLE users_by_country (
  country text,
  user_id uuid,
  name text,
  PRIMARY KEY (country, user_id)
);

-- 'country' is the partition key: its hash (token) decides
-- which node owns the row. Inspect token distribution:
SELECT token(country), country FROM users_by_country;

Follow-up Questions

• What is a virtual node (vnode) and why does Cassandra use many per host?
• How does the Murmur3Partitioner differ from the older RandomPartitioner?
• What happens to token ranges when a node is decommissioned?
• How does the coordinator node forward a request to replicas?
• Why does consistent hashing scale better than modulo-based sharding?