Performance degradation caused by aggr() functions after upgrading from 11.20 SR13 and below to 12.*

The Performance degradation after upgrade from 11.20 SR13 and below to 12.* is mainly noticeable in large applications with expressions containing the aggr() (aggregation) function.
This will result in increased response time when the user is navigating an app: after the selection of certain sheets or objects, after applying filters.

The bottom line of this issue is the lack of consistency by how the aggr function is used in this case. The resolution is to rewrite the aggrs so that the graph dimensions are included in the aggr expressions. Core Engine will detect that nice condition and parallelise the results, getting the performance back. An example to illustrate is the following:

The expression (indentation is by purpose to help readability)

sum(
Aggr(fabs(round(sum({<Last12Mos-=

{"-"},Flag={"RouteSales"}>} [Actual Gross Sale]))-round(sum(SIXM))),[Product Code])
)/round(sum({<Last12Mos-={"-"}

,Flag=

{"RouteSales"}>} [Actual Gross Sale]))

will not be parallelized since the dim of the graph is NODENAME11. To get this expression to be parallelized, the expression needs to be updated to 

sum(
Aggr(fabs(round(sum({<Last12Mos-={"-"},Flag={"RouteSales"}

>} [Actual Gross Sale]))-round(sum(SIXM))),[Product Code], NODENAME11
)/round(sum({<Last12Mos-=

{"-"},Flag={"RouteSales"}>} [Actual Gross Sale]))

Since we have a drill down group as dimension, the expression can be rewritten as:

sum(
Aggr(fabs(round(sum({<Last12Mos-={"-"}

,Flag=

{"RouteSales"}>} [Actual Gross Sale]))-round(sum(SIXM))),[Product Code],$(GetCurrentField('PRD-Drill-H4')))
)/round(sum({<Last12Mos-={"-"},Flag={"RouteSales"}

>} [Actual Gross Sale]))


Related articles:

• Data compression different in QlikView 12 than in QlikView 11
• Possible decrease in reload performance QlikView Server/Publisher 12.00 GA - Remedy