By default, EMR instances are configured with fewer gearboxes than cards. But reducers are not given additional memory, so it seems that they should have the same amount. (For example, super-large instances with a high processor speed have 7 cards, but only 2 reducers, but both converters and reducers are configured with a memory capacity of 512 MB).
Does anyone know why this is so, and is there a way to indicate how to use as many gearboxes as file cabinets?
EDIT: I had the wrong amount, it's 512 MB
( STDIN), , , . ( STDOUT) . , , .
, Hadoop MapReduce, EMR, , , , , jobtracker.
jobconf. - mapred.map.tasks mapred.reduce.tasks.