Like other MR jobs, you can specify the number of map tasks for the job via JobConf.setNumMapTasks(). However, this is only a hint and the actual number of spawned map tasks depends on the number of input splits. If you have set NumMapTasks=100 with 50 nodes, then the number of tasks in parallel per node is about 100/50 = 2 assuming the splits are evenly distributed across the nodes. ... View more