Hadoop Tuning

Small file size 
	1.MR - CombinedFileInputFormat
	  Hive - copy by fewer Reducers
	
	2.set input split size - block size - number of mappers( to bigger number)
			each mapper uses one jvm - fewer the mappers, fewer the jvms created and destroyed.
			if you have more mapper then smaller split size is better. - fewer mappers bigger size is better.
			
	
	3.allocating proper number of reducres
	  mapreduce.job.reducers - property
	  by default number of reducers determined by data size, does not have data statisitcs or cost based optimization.