Hi,
a) Packet size of a DTP:
Number specified here has a direct impact on memory consumption during extraction.
If the data is read in semantic groups, all the data records for a particular key have to be contained within the same package. For this reason, a package contains more data records than specified here
b) If the volume of the data is huge then you can create more number of DTPs and can perform selective loading at DTP level as well.
Also check for routines/formula at transformation level.
Hope it helps.
-- KRPK