I'm new to Qlik. We are evaluating QlikSense September 2017 release in an environment. The app is analyzing about 6 million rows. We see that even simple selections take 15 - 16 secs and the whole dashboard experience is not demonstration worthy. We believe that the expressions and calculations have been fine tuned to the extent possible. These have made some difference but we still need faster loading dashboards. Its just 1-2 users accessing the dashboards. We see that the application is able to max out the CPU when these elements are accessed.
The server has
Memory :380 GB
CPU : 2 x (R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz with 10 physical cores. I
HW platform : IBM blade Flex System x240 M5 Blade server.
Hyperthreading : We have tried disabling and enabling hyper threading. Not much difference seen.
Power Mgmt : Max Performance
Hardware prefetch is enabled.
Windows Performance settings are set to Background processes.
Page file is set to fixed size.
Is there anything else we could have a look at from the systems perspective?
Any input would be highly appreciated.
PS: We have downloaded the scalability tool but need to work on creating the right tests before we analyze the results.
I assume that there is something wrong with the datamodel and/or the expressions because even with much lesser hardware should the response-times a lot more faster. Please provide more informations about the datamodel und the used expressions.
Thanks Marcus. I unfortunately do not have this info as I'm only trying to find the best config for the server. I shared your concern with my team and they mentioned that they had already done everything they knew to make the data model better.
The application seems rather small and should be therefore run sufficiently on a local machine with the sense desktop maybe on an office-laptop. With this approach you should be able to exclude any config-issues with your used sense server, the OS and the network and similar stuff. If it runs near the server performance or even better the issue won't be the application ...
The problem was that the currency filter was missing and hence the drill down performed much more calculations than expected. Once the filter for a specific currency was enabled the charts performed well.
So the server settings were not the root cause for this reported issue.