DSXchange

What is the advantage of the option "recalculation" in the aggregator stage ?
Looks like it does the same thing as normal aggregation, is it only to avoid repartitioning and resorting on the same keys ?

Look at Pg 224 of Parallel Job Developer Guide for V8.5 give you a good example on the usage of Re-calculation.

Thanks for the reference but I don't have 8.5 mannuals yet, still running on 8.1. Is the same documented in 8.1 ?
I have definitely read an example in the guide of 8.1 (not sure if it's the same example you are referring to), it demonstrates how recalculation should be implmented but my question is "from what is documented it's clear that recalculation could be achieved using calculation mode in the following aggregator then what benefits does recalculation offer over to calculation mode"

"Re-calculation allows you to apply aggregate functions to a column that has already been summarized." - Parallel Job Developer's Guide

qt_ky wrote:"Re-calculation allows you to apply aggregate functions to a column that has already been summarized." - Parallel Job Developer's Guide

Agreed ! but why is there a need to provide an option of "recalculate" in the downstream aggregator stage when the same can be achieved using calculate option of the downstream aggregator stage ?

It's implied that it will save on processing. My imagination tells me that suppose you were tracking some total count and every day you had new data to add to the count, you could use recalculate to add the delta rather than calculating all days every day.

qt_ky wrote: My imagination tells me that suppose you were tracking some total count and every day you had new data to add to the count, you could use recalculate to add the delta rather than calculating all days every day.

Your imagination seems good. But i am not able to imagine how we can implement this in real time using recalculation.

Think about the possibility of updating shrunken (summarised) fact tables.