What takes less memory?

[richard_simpson]

I suspect that there isn't an easy answer to this and it may depend on all sorts of other variables... but what do people think?

Is a Star-schema model with 50 million facts and one value field better or worse than the same model but with values merged onto facts so that there are 10 million records with 5 value fields? By value I mean numbers... not text.

This Board is called Qlik Sense App Development but I think this question can apply to both Qlik Sense and QlikView. Or can it? 

[Anil_Babu_Samineni]

It is not recommended to have all facts into single application for both View and Sense since it has QIX engine

Anyway, as a workaround you can create multiple layers with different fact and club them in to QVD then use this Fact QVD in presentation layer

[richard_simpson]

I'm sorry but I'm afraid that I don't understand your reply to my question 😕

I was talking about one application (QS or QV) and using a single Value field in one large Fact table. Each record would also have a flag i.e. flag_ValueIsSales (set to 1 or 0), flag_ValueIsCost (set to 1 or 0) so that chart expressions can simply say: Sum(Value * flag_ValueIsSales) to calculate Sales and Sum(Value * flag_ValueIsCost) to calculate Cost. The idea being to optimise memory footprint of the app because the flag fields only have 2 different values (tiny memory usage when Qlik stores unique values) and all the different values are held in one field (Value). 

I know that this is an efficient way to build an app.... but I'm wondering at what point the disadvantage of having a very large number of rows outweighs the advantage of storing all the different numeric values in one field?

I hope that makes more sense?

I'm not sure what you mean by "not recommended to have all facts into single application" or "create multiple layers with different fact and club them in to QVD"?

[Ksrinivasan]

hi,

that is the beauty of qvd form data of storage and retrieve.

below indication are the  cause  if, the very large number of rows outweigh different numeric values in one field.

1. processing or result of visualization time may increase,

2. QS services will stop while loading script

3. sheet performance getting slow for other users 

but developer must know about data structuring from raw data and

how data processing happing in the APP, like indexing method,

moral is vehicle efficiency is in drives hand (who knows function of gear, clutch and break).

ksrinivasan

[marcus_sommer]

Hmmm, it depends and the file-size and memory consumption is only one aspect if an application fulfilled the requirements or not. I think in your case (multiplying  a few flags with values) I would tend to the shorter and wider fact-table. A similar discussion could you find here:

Data-Model-Question-Fat-or-Thin-Fact-table 

- Marcus