Qlik Sense engine memory usage parameters tuning

[Dumbest_in_the_room]

Hi,

We have 64GB single site environment with 20 users, 5 concurrent users usually, and experience a fast saturation of RAM after 4 days of services restart.

We ed potential areas of tuning in following engine parameters, currently at deafault levels:

- app cache time
- min memory usage
- max memory usage
- export limit (bytes)
- export limit time  

We think app cache time is the prime target, and would like to diminish it from 8 hours (default) to, say 2 hours. Is there anyone who tried such levels and can tell if there are side effects?

Are there side effects in lowering min memory usage from 70% to, say 30%?

Thank you

[Bastien_Laugiero · Accepted solution]

Hello, 

Thank you for your question.

So the side effect of lowering the app cache timeout to two hours is that any applications without any user activity will be unloaded from memory after 2 hours meaning that the application will need to be re-loaded into memory which takes extra time on opening. So depending on how often users are accessing applications during their working day, this parameter could be valuable.

 Regarding the Min memory usage, I wouldn't recommend to decrease it. Once reached, the Qlik Sense Engine will start trying to minimize its allocation of memory meaning releasing information from cache. Setting it up to 30 % means that the engine will reach this value very early and therefore have a negative impact for the users as well as CPU on the Qlik Sense server. 

One thing you may want to look at is how much memory your applications are consuming once loaded into memory. Perhaps you have a few applications consuming a large portion of the memory and some optimizations in those apps will be a great win. Here is an article that could help you determining that: https://community.qlik.com/t5/Support-Knowledge-Base/How-to-enable-Qlik-Sense-QIX-performance-loggin...

Finally, since you are running on a single node environment, it means that the reload tasks are also consuming resources. You may want to check the reload task schedule to avoid as much as possible any reload during working hours. (also applicable for the developer reloading their apps manually from the hub)

Hope this helps!

[Dumbest_in_the_room]

Thank you Bastien,

I think you nailed some key points here:

1) multi node site would isolate reload RAM usage from user RAM usage

2) single apps (and sometimes single objects in apps) follow a Pareto rule: Few are responsible for the greatest part of the issue.

Thank you also for the link to the Telemetry, we will most definitely activate it!

Kindest Regards