I have access to a small group of heterogenous servers:
A has four cores, 16gb
B : four such machines , eight Cores ,16gb
C: two, each with 16 Cores and 128gb
Each of these will be tasktrackers and datanodes.
What should be the appropiate values for mapred.max.map/reduce.tasks
For A and B I have 6 and 4 respectively( yes a bit much for A), what
is recommended for C?
Thanks and regards