is there a tuning parameters or configurations for Hadoop to utilize the underlying hardware capabilities (like multicore, processing threading etc ) or it is totally agnostic and abstracted from the underlying hardware?
Best Regards, Mohamed El-Refaey
You can control the number of tasks per task tracker via mapred.tasktracker.map.tasks.maximum property and mapred.tasktracker.reduce.tasks.maximum . If your machines has more core, you can give a happy number. Recommendation is to give number between 1XCores and 2XCores.