Exploring HDFS architecture
The HDFS architecture is based on master and slave patterns. NameNode is a master node and all DataNodes are SlaveNodes. Following are some important points to be noted about these two nodes.
Defining NameNode
The NameNode is a master node of all DataNodes in the Hadoop cluster. It stores only the metadata of files and directories stored in the form of a tree. The important point is NameNode never stores any other data other than metadata. NameNode keeps track of all data written to DataNodes in the form of blocks. The default block size is 256 MB (which is configurable). Without the NameNode, the data on the DataNodes filesystem cannot be read. The metadata is stored locally on the NameNode using two files—filesystem namespace image file, FSImage, and edit logs. FSImage is the snapshot of the filesystem from the start of the NameNode edit logs—all the changes of the filesystem since the NameNode started, when the NameNode starts, it reads FSImage file and edits...
