Qliksense Mashup initial load is very very slow

[wangxusq]

Hi all,

I created a mashup and the path is the default path.

But i found initial load is very very  slow. I also refer to some articals in the community.But no work.

So what can I do?

Or I want to add the loading picture,when all objects are loaded,reports begin to display.

but the problem is  when these objects have loaded?

Thanks

[ErikWetterberg]

Hi,

Why is it slow? Is opening the app in the built-in client also slow? Have you tried using the console to find out what it is that takes time, is it the openApp call or the getObject? Are you loading many objects on startup?

Erik Wetterberg

[wangxusq]

The avg time is 10s,my clients said it is Bad experience。

Every time the title appears,but the report is slow,like this:

I analysis the performance,it spends much time on script(qlik.js).

So have idea to impove this?

[paulcalvet]

Hi Scott,

I have the same problem, initial load is very slow (between 10 and 20s).

Did you resolve it ?

Thanks

Paul

[balabhaskarqlik]

If you resolved the issue, post the solution to let others know.

[jcamps]

n case it helps you,

We had the same problem with a very complex document, and a not very big size.

I run several tests with a new mashup by progressively decomposing the document into simpler and simpler versions:

First time the document loads ever is always slower (disk to RAM loading, I guess). I did not take into account that.

Also, note that having another tab pointing to the document in the same machine seems to do strange caching things.  Even if they are in different incognito browsers or inside iframes or whatever. That includes the document in the hub, the mashup editor, and obviously having the mashup open at the same time. So do close all tabs before you get mad with test results.

In my case, I progressively removed:

 - (A) Mashup complexity, by creating a simple mashup with the dev hub and just one object created: improved about -50% load time. (from about 25 to about 15 seconds)

 - (B) Many sheets (a few with several dozens of fields): No impact (in my case, about 15 seconds)

 - (C) All sheets (leaving just a couple of objects to load in the mahup): No impact

 - (D) Recreated the document from scratch with that single object, by binary-loading from the other document (so same data set but zero inherited objects): about 2 seconds loading time (!!!) (.. after the first load with takes about 8-9 seconds).

So, data does not seem to take that long to load. I then started removing all "my stuff" from the (C) document :

 - Removed all measures --> No impact

 - Removed all dimensions and objects in the catalogue --> No impact

 - Removed all alternate sets --> ( I had around 8-10); This seems to have a certain impact (about 20-30%)  but far from the 2 seconds

 - Removed all variables --> 

Note: to remove the variables I did the following:

            a) removing all existing script and changing it into a simple binary load from a copy of itself. That let go of the annoying behaviour that script-created variables cannot be deleted, since they all got "separated" from the script.

            b) creating a Selenium script to automate removal with Katalon recorder), since Qlik Sense does not have multiselect to remove variables from the hub.

As the humongous amount of variables got lower, load times went steadily down, from 14 to 2 seconds.

I have to say none of the variables had the equal sign - that should mean they are not evaluated. And when I say many,I mean many......  in the order of several hundred. We use a variable to store each metric's expression. Each one has different "variations" (py, ytd, vs, ....) and we have a large amount of metrics. Hence the number.

But still.... the impact surprised me.

I hope this experience can help others.