Can you describe difference in performance when using 3.x emulated and migrated DS and old and new concept IS?
It is very important for us because we start new project on 2004s, and should choose Data flow concept.
Loads of large amount of data are supposed.
In general, I would choose the new data flow concept as it has tons of benefits over the old concept. The performance while loading, in general, is more parallelized with the new data flow, so you will generally get a ton better performance. You can run some tests if you need numbers, but I think you'll find the new data flow concept a ton better to work with.