While it is possible to deploy all of HDP on a single host (single node deployment), this is appropriate only for initial evaluation. In general you should use at least three hosts: one master host and two slaves.
Also, see: Typical Hadoop Cluster.
Single node deployment uses separate processes for each of the Hadoop services (NameNode, DataNode, JobTracker, TaskTracker) on a single machine. However, this Hadoop cluster is not truly distributed, because no processing or data storage is performed on remote nodes.
