hm ok thank you!
I have 60 Map Tasks for 15GB and 427 for 100GB. And 16 Reducers and 111 for 100GB. Which i think is way too much but it seems like hive is not doing that partial aggregations within the big data since the reduce phase is really fast like seconds when i use combiner in my Mapreduce but takes forever with hive.
With Mapreduce i have like 32 Mappers and 1 Reducer for 15Gb and 212 Mappers and 1 reducer for 100Gb, i increased block size to 512MB because its faster.