The NameNode component of HDFS was the central point of failure in the initial versions of Hadoop. In the later versions, a secondary NameNode was introduced as a backup for the primary NameNode. Until Hadoop 2.X, the NameNode component could only handle a single namespace, making it less scalable and difficult to isolate in a multitenant HDFS environment. Scalability and isolation were the two most desired requirements for Hadoop enterprise deployments. Most organizations shared infrastructure among their different teams with varying degrees of availability and authorization aspirations.

HDFS Federation is a feature that enables Hadoop to have multiple namespaces, making it easy to use for shared cluster scenarios. This feature brings about a separation between the storage and namespace management. Similar to YARN, this separation helps onboard other applications and use cases on to HDFS, making Hadoop move away from a MapReduce-only platform to a more generic...