Architecture of HDFS Federation

The crux of the HDFS Federation feature is that it allows for multiple NameNodes to run on a cluster. These NameNodes are independent and do not have any dependency on each other. However, the DataNodes are shared between all the NameNodes in the system. The NameNodes are said to be federated because they can be run independently without coordination.

Each DataNode sends heartbeats and block report information to all the NameNodes in the cluster. DataNodes also receive instructions from all the NameNodes. They are the common shared storage resource in the cluster and still run on commodity hardware. However, they cater to different NameNodes, and in turn, facilitate different Namespaces. These independent Namespaces ...

Get Hadoop: Data Processing and Modelling now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.