Be advised that table splits are done 1:1 with the number
of served regions while file splits are done in a totally
different manner. Can you confirm (or not) that your input
table has split? If so, how many regions are there? When
you run your job using files, how many file splits /
mappers are allocated? When you run your job using an input
table, how many mappers are allocated for that? Likewise
for the output side: How many reducers? How many output
files? How many regions in the output table?
Thanks,
- Andy
