topic Re: Metric Performance issue in QlikView

Metric Performance issue

Chanty4u — Fri, 22 Apr 2016 11:51:55 GMT

We are trying to create a metric-- ignoring all the filters in the layout except for two FIELD1,FIELD2

we are able to get the correct data. Using the following expression

Count(Distinct TOTAL<FIELD1,FIELD2>{1<FIELD1=$::FIELD1,FIELD2=$::FIELD2>}USERS)

But we run into serious performance issues.

There are two tables TABLEA contains FIELD1 and has 15million rows

TABLEB contains FIELD2 and has 1million rows

Both these tables are joined on USERS id. Please suggest an alternatice approach to get the same result with better

performance

Thanks in advance

Re: Metric Performance issue

MK_QSL — Fri, 22 Apr 2016 12:01:29 GMT

Not sure if this would help you or not...

Ignore all selections except some specific fields using Set Analysis

Re: Metric Performance issue

marcus_sommer — Fri, 22 Apr 2016 12:04:42 GMT

I think you should consider to change your datamodel and transfer both fields into a single table then otherwise will be the virtual table which needs to be created very large, see also: Logical Inference and Aggregations.

- Marcus

Re: Metric Performance issue

swuehl — Fri, 22 Apr 2016 12:05:27 GMT

In which context are you using this expression? Chart (type)? Dimensions?

Re: Metric Performance issue

sunny_talwar — Fri, 22 Apr 2016 12:11:14 GMT

I am assuming you have both Field1 and Field2 as dimensions in one table which might be creating a massively huge table which might be causing the issue

May be join the two tables in the back end to get a better performance on the front end

Re: Metric Performance issue

Chanty4u — Fri, 22 Apr 2016 12:18:16 GMT

Hi Swuehl,

We are using this expression in a measure which is used in a pivot table.

Re: Metric Performance issue

Chanty4u — Fri, 22 Apr 2016 12:26:11 GMT

Hi Sunny,

We tried combining the Table 1 and Table 2 which has resulted in a huge data set. Which lead to considerably long load time.

We would really like to keep the backend tables as is,

Is there any alternate expression that we could use in the measure to improve the performance but get the same result set.

Re: Metric Performance issue

sunny_talwar — Fri, 22 Apr 2016 12:29:51 GMT

I guess it comes down to a choice between back end load time vs. front end performance