It's not complete nor authoritative, but http://www.cloudera.com/blog/2009/08/14/hadoop-default-ports-quick-reference/ has a table of what ports which daemons have open, and what protocol they use to communicate. Most communication is Hadoop IPC, but there's some raw socket stuff for the datanodes, and there's some HTTP too.
-- Philip On Wed, Aug 19, 2009 at 12:14 PM, B. X.<bxi...@gmail.com> wrote: > greetings, > > Is there a summary somewhere regarding the different APIs used > between JVM processes in Hadoop to communicate with each other? I > found at least the following: > - using SocketChannel with apache/hadoop/ipc/[Server|Client] apis; > - using IO streams customized in > apache/hadoop/hdfs/server/datanode/Block[Sender|Receiver]; > > Thanks, > -Bin >
