Refer http://sortbenchmark.org/
Three version of program are made.
input file is:
One is local one, where program uses external sort to sort the data and uses im-memory data-stucture.
Hadoop-MR and Spark uses 16 nodes of d2.xlarge on aws to sort the 1 TB data.
For deployment use the script in hadoopsetup repo while follow manual steps for Spark.