Considering the PostgreSQL and required performance tuning - I have a feeling that there is never a good solution to that kind of problem - you either stick to default settings for each software being tested and get remarks like this, that this or that should have been optimized/changed/set to different value or you can optimize the settings but this leads to comments that benchmark is biased because it settings for X were optimized and settings for Y were not (or something similiar).