Cassandra is masterless—there are no primary and secondary nodes/replicas. A client can connect with any node in the Cassandra cluster. Once connected, the node acts as the coordinator and drives the required interactions with the rest of the cluster on behalf of the client.
On a single node, Cassandra uses log-structured merge trees to store data. The life cycle of a write is as simple as updating an in-memory table of row/column values and writing to a commit log (so that durability is retained):
Periodically, the in-memory table (called memtable) is flushed to disk and the data is converted into SSTable format—where the keys ...
