Harsh J <qwertyman...@gmail.com> wrote: > The source is your friend.
In particular, start with JobTracker.java: initJob() for Job and TaskInProgress creation/init, and heartbeat() for Task creation. TaskTracker.java, JobInProgress.java, TaskInProgress.java, *Task.java, *TaskRunner.java, *TaskController.java, and JvmManager.java are the other core MR files, I think. Internal communication protocols include InterTrackerProtocol (JT-TT), TaskUmbilicalProtocol (TT-Task), and HTTP/Jetty (MapTask-ReduceTask and others, I believe). There are probably more of them, but those are the ones I've come across so far. I haven't dug into HDFS. Greg
