I am going through hadoop definitive guide, where it clearly explains about input splits. It goes like
Input splits doesn’t contain actual data, rath
HDFS block size is an exact number but Input split size is based on our data logic which may be a little different with the configured number