Skip to main content
Announcements
Qlik Connect 2024! Seize endless possibilities! LEARN MORE
cancel
Showing results for 
Search instead for 
Did you mean: 
Kalkumar
Partner - Contributor III
Partner - Contributor III

Qliksense Slowness and High RAM Utilization

Hello All, 

I'm experiencing very slow dashboard opening, selections in it.

1)We have 3 node architecture , Central, RIM and a dedicated PostgreSQL. 

2)RAM is 32 GB and CPU is 8

3)Data model is simple (not very complex)

4)Used App Meta data monitoring app and didn't find any specific item of concern.

5)Moved all the IF statement to Pick and Match and where ever possible to the script level.

6)Using Optimized QVD load.

7)QVF size is 480MB 

8)Number of records is 700K

9)No Syntactic key 

10)Few Data island (very less data and just 4 to 5 columns in it)

11)We are using few VizLib extension. But when we checked with VizLib , they stated the QVF is faster in their env and doesn't provided much help. 

Sometimes one page/sheet take 9 - 12 seconds to open and at times it take 25  - 30 seconds to open a single sheet/page.

Any suggestions ?

Labels (5)
1 Solution

Accepted Solutions
Albert_Candelario

Hello @Kalkumar,

Thanks for posting.

A 3 nodes environment with nodes of 32GB and 8 cores, seems quite small, of course will depend on the number of users, apps and so on.

I have few recommendations to make, move to the patches mentioned here to ensure you enjoy the latest performance improvement when opening apps:

Qlik Sense Enterprise on Windows: Improve app open... - Qlik Community - 2134472

Check you are not using virtual memory, so the disk instead of RAM, since 32 GB as mentioned seems pretty small.

Have a look at:

STT - Log File Analysis App for Troubleshooting Ql... - Qlik Community - 2091350

Last at not least, to have a healthy environment review this resource:

Home | Playbook - qs-admin-playbook (qlik-poc.com)

Cheers,

Albert

Please, remember to mark the thread as solved once getting the correct answer

View solution in original post

7 Replies
rwunderlich
Partner Ambassador/MVP
Partner Ambassador/MVP

I would recommend analyzing the app with QSDA Pro

Also, can you view the WIndows Task Manager on the server during this slowness to see if the server is CPU busy or RAM busy?

-Rob
http://www.easyqlik.com
http://masterssummit.com
http://qlikviewcookbook.com

 

Kalkumar
Partner - Contributor III
Partner - Contributor III
Author

@rwunderlich  Looks like QSDA Pro is not a Qlik product, but a seems like 3rd party app/tool. 
Don't think firm security policy would allow it's installation. 

rwunderlich
Partner Ambassador/MVP
Partner Ambassador/MVP

True, it's a third party product from Motio, Inc, a respected vendor of BI tools. I'm the author of QSDA Pro. 

-Rob

marcus_sommer

Pick(match()) is much better usable as if-loops but from a performance point of view it's quite the same. Therefore try to replace it completely. Also if there any aggr() constructs in use.

Further make sure that the fields from the island-tables aren't associated on a dimensional context within the objects because this creates cartesian products.

Kalkumar
Partner - Contributor III
Partner - Contributor III
Author

We are not using aggr() in any of the measure. 
Could you please elaborate more on "island-tables aren't associated on a dimensional context within the objects because this creates cartesian products".

There is no association between island-table with the other tables of the data model. 

marcus_sommer

That such tables aren't connected to other tables within the data-model makes them to island-tables.

At this point it's ok. and not a problem. But if you use a field from the island together with fields from the regular data-model within an object it could cause performance issues, for example by using a period-field from the data-model as dimension and a category-information from the island-table - which are matched together by any conditionally logic.

In such case it creates at first a cartesian product of all possible field-combination and afterwards comes the match respectively reduction in regard to the applied conditions and the last and least will be the aggregation against this dimensional context.

480 MB for only 700 k of records is rather large and hints for quite wide tables and/or larger string-content in some fields and/or (several) fields with a high cardinality respectively many distinct field-values like record-id's or timestamps. Nevertheless it's a rather small data-set and opening sheets and/or applying selections should be performed without waiting for seconds. Therefore something seems to be quite unsuitable within your application.

Albert_Candelario

Hello @Kalkumar,

Thanks for posting.

A 3 nodes environment with nodes of 32GB and 8 cores, seems quite small, of course will depend on the number of users, apps and so on.

I have few recommendations to make, move to the patches mentioned here to ensure you enjoy the latest performance improvement when opening apps:

Qlik Sense Enterprise on Windows: Improve app open... - Qlik Community - 2134472

Check you are not using virtual memory, so the disk instead of RAM, since 32 GB as mentioned seems pretty small.

Have a look at:

STT - Log File Analysis App for Troubleshooting Ql... - Qlik Community - 2091350

Last at not least, to have a healthy environment review this resource:

Home | Playbook - qs-admin-playbook (qlik-poc.com)

Cheers,

Albert

Please, remember to mark the thread as solved once getting the correct answer