topic Re: best practices of data modeling in QlikView

best practices of data modeling

maxsheva — Tue, 24 Jul 2018 11:44:10 GMT

Hi All,

What are the best practices of data modeling if there only one data file includes both measures and dimensions?

The number of columns are 24 but rows are not limited as new data will be loading every day.

Could it be one table or need to divide them by Fact and Dimensions?

I remember one table can take more CPU resources where Star as well as Snowflake schemas affect on RAM.

Any suggestion will be much appreciated.

Re: best practices of data modeling

PabloTrevisan — Tue, 24 Jul 2018 11:55:19 GMT

Hi Max,

Look this image please:
It's in portuguese ok?

If you need a data model, prefer the star schema, he is the best in most of cases.

Sometimes it is not possible to use it, so we chose Snow Flake or Simple Table.

Re: best practices of data modeling

tresB — Tue, 24 Jul 2018 11:58:45 GMT

In short,

Generally, with single table your front-end performance would be better, however, trade off comes in script load, application size; load time could be longer. And with other approach, the effect would be reverse.

Re: best practices of data modeling

maxsheva — Tue, 24 Jul 2018 12:49:20 GMT

Pablo, I have already found that example in English Regarding Star schema to a single table | Qlik Community

However it cannot give me right decision.

tresesco thanks for suggesting. Do you know where to find test evidence about better front-end performance when to use single table? Why do you think picture above shows worst RAM consumption in a using of single table?

Re: best practices of data modeling

tresB — Thu, 26 Jul 2018 06:11:09 GMT

I don't know about the test evidence source. However, single table taking more RAM makes sense to me. Let me try explain why I feel so.

A single table means your table is de-normalized upto maximum extent, i.e. - data redundancy at it's best (actually worst for us). Though qlik uses it's symbol tables using bit-stuffed pointers to reduce this redundancy at the back-end (disk), when expanded in the RAM the same redundancy prevails and RAM is consumed accordingly.

Re: best practices of data modeling

vipin_mishra479 — Thu, 26 Jul 2018 06:27:16 GMT

As Pablo has mention Start schema is the best response in data model Even Qlik also suggest the the Start Schema in his Best Practise for data model. Even you can use optimised qvd. And use specific data.

If you have multiple fact table with diffrent granuality level data you can use Link table.

If you have same Granuality level data you can preffer Concatination.

Re: best practices of data modeling

maxsheva — Thu, 26 Jul 2018 08:38:29 GMT

Hi Vipin, I believe you mean Star Schema instead Start

Thanks all for help.

Wish you the best!

Yevhenii