What is HDFS Federation?How it helps namenode in scaling ?
MapR's architecture is different from HDFS, and one of the most significant differences is that we do not use a namenmode - instead, the metadata is distributed across the cluster along with the data. The namenode limits scalability, since all of the cluster's metadata must fit in memory on the dedicated namenode.
HDFS Federation addressed this issue by allowing you to have multiple namenodes - so when your cluster is getting to its size limit, you need to add another machine to act as a namenode.
That's better than being stuck with a single namenode, but MapR's distributed metadata eliminates the problem entirely.
Retrieving data ...