QVD load

[#1]

Hello

I have loaded data into my qlikview and i have above 70 mio rows. This takes quite long time to reload everytime, and to develop in, because i need to save all the time.

What is the best practice to do?

Is it A:

Data.qvw and store this into QVD.

QVDLoad.qvw -> Load data from QVDs and  Design here?

Or is it B:

Data.qvw and store this into QVD

QVDLoad.qvw -> Load Data and make the nessecary scriptning

Design.qvw -> Binary load from QVDload.qvw and design in here.

Or what?

[#2]

Data.qvw and store this into QVD.

QVDLoad.qvw -> Load data from QVDs (Only required period)and  Design here..

Try to put less calculation in your UI.as much as do it in ETL time.

[Joaquin_Lazaro]

Hello Thomas Jensen

I suggest you:

1) DataHistoric.qvw ... create qvds with data from first day until 2012/12/31

2) DataAnual.qvw ... create qvds with data from 2013/01/01 until 2013/xx/31 (xx=previous month)

3) DataCurrent.qvw ... create qvds with data from 2013/zz/01 day until today (zz=previous month)

4) App.qvw ... load qvds* and contains all the charts and analysis insigths

Hope help you

Joaquín

[brijesh1991]

Binary Load will be the best. ..

Also if you will make changes in Data Model file, then take partial load using add/replace function at scripting level. . .

[kaushiknsolanki · Accepted solution]

HI,

     Both the approaches are good.

     But there are few things which you need to make sure to improve the performance.

     1. DataModel.

     2. Move all your script calculation to Qvd Generation, like any if statements or any Composite key generation.

     3. Dont use if statements in charts.

     and many more.

     For more information have a look at the link below.

     http://community.qlik.com/docs/DOC-1881

     http://community.qlik.com/docs/DOC-1882

Regards,

Kaushik Solanki

[#6]

I already did as much ETL calc as i can. And i need all period

[#7]

Uhm. I see your point. But the fact is that i have some data, which is not dependent of a date. So i need ALL data.

[#8]

Hello,

1. By this you mean (Have my datamodel in place) My DataModel is well considered, so i dont think i can gain much more performance or less rows here.

2. So all my scriptning in QVD. Thanks - That was also what I saw as the "best way"

3. Trying not to

I think im going with the binary load. It seems faster, and loads in few minutes, where the old way took 30 mins

[Joaquin_Lazaro]

Normally the fact table depends on date.

With qvd you'll also have all the data.

Try save all the data depending out of your document in qvds and load the app from qvds and mesaures/compare the duration of load script.

[#10]

Yes! But some of my data is not date depended. Maybe 60mio of my rows is not date depended