Speed Up my QlikSense

[#1]

Dear Community,

i often saw apps and marketing-campains, where the user was able to load Data into the visualisation within millisecounds or secounds.

I was creating an App with a table with at least 3 million line and about 5 dimensions and another table with about 20 thousand lines and about 4 or 5 other dimensions and a shared key for both tables.

At first i joined the tables to one large table, but on the current version i released this connection and had this 2 tables.

So i tried different ways, but hadnt found the best practice.

When i now loading my app, on a server, it could take up to 1 minute before the visualisation is complete and about 20 secounds after setting a filter.

What is the best way to increase the performance?

The data are stored in two different qvd-files.

I think 3 million line only a few dimensions and attributes and one single join

How can i speed up the loading process without reducing data?

Thank you for your time and help.

[ganeshreddy]

Hi Alexander,

Your data seems not that big, please double check your join case.

However try to do QVD optimized load, you can see drastic improvement in performance. If you use any transformations or where clauses while loading it won't comes under optimized load, so try. to avoid them.

If you still need to apply transformations and where clauses while loading- use temp tables, resident tables, where exist clauses to achieve the need.

Thanks,

Ganesh

[ArnadoSandoval]

Hi Alexander,

Your application is based on two QVDs the should be joined by a single common column (you did not elaborate about their join), these are some performance enhancement tips for you:

• Memory, for the size of your application, 8 GB is good, 16 GB is better, 32 GB fantastic.
• If the two tables are joined on a single column, and this column is not numeric, implement the AUTONUMBER function.
• Your solution should not have synthetic keys, with data from two QVDs this is unlikely, but anyhow, something to check.
• There are more tips on John Whiterspoon reply here

[#4]

For information only:

there is an bug with the aggr-function.

Aggr funciton in combination with set Analysis is very "memory hungry"

The used dataspace is dramaticaly exploding, when using this function on large, diverse data sets.

Preaggregation could solve this problem