DSXchange

Hi,
I currently have a design where I need to Sort and Aggr.
Do you recomment a Sort in AGGR stage (or) a SORT stage followed by a AGGR stage. If so, why?
Thanks!

I would always sort before the aggregator. The sort method to use really does depend upon your incoming data (i.e. coming from a table you can use the Database's mechanism, a flat file might be better sorted outside of DS, etc.).

I agree with ArndW. The agg stage is best doing its job alone.

Wouldnt it be better to use the parallelism mechanism on the SORT in DS instead of a flat file sort outside of DS.
My understanding is a sort of DS is not utilizing the PX mechanism. Im assuming partitioning and sorting in the SORT stage would be faster.
Thanks! Pl. reply!

Do you need to sort the entire DataSet? Or just the data on each partition? In the latter case, the PX sort may well be faster.