> Can anyone tell me the purpose of sockets used in hadoop. I mean
> for which purpose hadoop uses sockets? are they used by tasktracker and
> jobtrackers ? or used by namenode and datanode ?
All communication between Hadoop masters (JobTracker, NameNode),
slaves (TaskTracker, DataNode) and clients happens using a custom RPC
built over sockets.
> Secondly, if hadoop has to use sockets does it uses the socket factory
> classes given in org/apache/hadoop/net directory?