Memory of Hidden objects in server

[sundarakumar]

Hi All,

Will a Hidden object consume server memory while the user access the particular sheet, where this hidden object is present? or is it like a minimised object.

This hidden object a Pivot table which will be shown only on a filter selection. The hide condition is set in layout tab.

Can someone pls help with this?

-Sundar

[jonathandienst · Accepted solution]

Hi

I doubt that memory consumption is causing poor performance in your dashboard, unless your server has too little RAM for your application.

I would look at the structure of your model -

     Is your data model a well formed star schema, or is it more complicated?

     Are the tables associated using a single key, or a simple composite key?

     Are there a lot of rows in tables that are not associated with any other tables?

And the expressions -

     Do you use calculated dimensions (especially if these could be replaced by derived fields during load)?

     Do you use sum(if()) structures or similar structures requiring massive joins

     Are your expressions very complex (Pick(Match()) or deeply nested If() statements, many terms...

Any of these can cause performance issues, especially when the fact table(s) contain millions of rows.

HTH

Jonathan

[sundarakumar]

Hi Jonathan,

Thanks for the suggestions.

Model Structure.

1. it is kind of a fact constellation where I Have 2 main fact tables linked to each other. The diemnsions get linked to either of the facts. There is a main dimension table which gets linked to the facts with the key that is used to link the facts together, this creates a node of 3 tables (2 fact and 1 diemnsion)

2. Association is with a composite key which has four fields concatenated. it is numeric

3. there are only few island tables, they have less row count.

Expression:

1. there is no calcuated dimension used.

2. no if statements

3. mostly sum() and only(). we use set analysis and if in some cases. but no a very complicated expression

The most important thing is fact table has very less number of columns not even a millon.

Server RAM - 16gb

application size  -  6 mb

This is y am worried abt the performance. This should be good. do u think the data model is the issue out there?

or the hidded pivot?

Please advise...

-Sundar

[jonathandienst]

Sounds OK, except for the island tables. Are these used in the objects that perform poorly. When there is no association, QV has to perform a cartesian join to perform calculations which use fields from the main fact and the island table. This can cause a large spike in memory and affect performance badly.

Are the key field values unique in the dimension and in the 2 fact tables?

The model size on disk is not a reflection of how much RAM it requires. The compressed qvw expands on loading (possibly to between 1 and 2GB in your case), and will increase further if joins are necessary to perform the calculations. Check using task manager.

HTH

Jonathan