Why does publishing a Qlik Sense app affect it's performance?

[stascher]

I had a client complaining of slow Qlik Sense application response times whenever he made any filter selections. It would take about 30 seconds for the two charts on a sheet he was viewing to render. He was the only user accessing any Qlik apps in his stream and the app was not in the process of reloading.   I went in through the Qlik Hub, accessed the same app and observed the same poor/unacceptable responsiveness.  

I then duplicated the app via QMC into my workstream. Opened the app, performed the same selections on the same sheets (repeatedly) and responsiveness was nearly instantaneous.  Same app, same data, same sheets, same selections.. vastly different response times. 

So my question is, what is is about being published that can degrade an application's behavior? And more importantly, what can I do about it? 

Thanks,

Steve

[Bastien_Laugiero]

Hello, 

I see a couple of reasons why the published application may be slower (but perhaps not to the extent you are describing)

• User Object: When duplicating an application, only the base object are being duplicated and not the objects created by users. So if the application contains a lots of user object, then this may explain why the duplicated version is faster.
• Security rule evaluation: Depending on the setup and complexity of your security rules the evaluation could take more or less time depending on which application you are working on. 

Have you tried to publish your duplicated app under a new name and see if it provided the level of performance you are expecting? 

Hope this helps! 

[tresesco]

In addition to @Bastien_Laugiero 's points, there could be another possible reason comes into my mind:

        - if your deployment architecture is so that work stream apps are handled by one node and the published apps are loaded to another node, then the node capacity and/or usage might be one reason.

[stascher · Accepted solution]

Thanks for the responses. I'm sure they're all valid points. Turns out that in this case there was a 3rd-party extension that was the root cause.