With No NameNode HA™, MapR scales linearly with the number of nodes, providing unlimited file support. Simply add nodes to increase the number of files supported.
In other Apache Hadoop distributions, the NameNode runs on a single server, even for very large clusters which creates several problems. Even with an exceptionally powerful server, the number of files is limited to only about 70 million files. To attempt to work around this issue, many large Hadoop sites actively run Hadoop jobs to walk through the cluster and concatenate files — amounting to a significant percentage of their daily jobs and wasting both resources and money.
MapR scales to support an unlimited number of files.
|Apache Hadoop||MapR Distribution for Hadoop|
|Data in Cluster||20PB||1000PB|
|Number of Files||70 Million||1 Trillion+|
|Number of Volumes||-||100,000|
|Performance||1x||2x - 5x|