LeoScott :
When you run your map-reduce from eclipse (i.e. without building .jar and directly running by clicking run program) then you are actually running the map-reduce in pseudo mode(mostly used for testing purpose). \n\nTo run a program in pseudo mode, you don't really need to have hadoop installed. You just need hadoop jar files.\n\nBut when you use hadoop-mapreduce-example-2.6.0.jar (I.e. you write command to run it on terminal) you are actually running the mapreduce on live cluster (Hadoop). For this you need your Hadoop cluster up and running.",
2017-01-16T07:15:58
SMA :
There is no difference. Just that when your input is huge and you have mapper/reducer running on multiple nodes, you would see drastic performance improvement as now the word counting will be done in parallel across different machines.",
2015-02-07T15:24:52