You set the replication factor during creation of the keyspace. So if you use (and plan for the future on using) a single data center set-up, you create the keyspace using cqlsh like so
CREATE KEYSPACE "Excalibur"
WITH REPLICATION = {'class' : 'SimpleStrategy', 'replication_factor' : 3};
Check out the documentation regarding the create keyspace. How this is handled internally is related to the snitch definition of the cluster and a strategy option defined per keyspace. In the case of the SimpleStrategy above, this simply assumes a ring topology of your cluster and places the data clockwise in that ring (see this).
Regarding consistency, you can very different levels of consistency for write and read operations in your client/driver during each operation:
Cassandra extends the concept of eventual consistency by offering tunable consistency―for any given read or write operation, the client application decides how consistent the requested data should be.
Read the doc
If you use Java in your clients, and the DatatStax Java driver, you can set the consistency level using
QueryOptions.setConsistencyLevel(ConsistencyLevel consistencyLevel)
"One" is the default setting.
Hope that helps