Skip to main content
cancel
Showing results for 
Search instead for 
Did you mean: 
Not applicable

cache warmer?

Happy M-day, everyone in the USA.

So, I've notice one bit of thing that is helping the slow browsing issue.  When I open my dashboard from the browser, for the first time, it is slow.  using a different browser, later, turn out to be faster.

It reminds me of the MS product that, they have a way to load the data to be browse into memory.  Someone call that technique a cache warming; liked the data is cached and all warm-up ready to go.

is there anything liked that for QV11?

I have a App server, web server, and a Dist server, then clients.

29 Replies
alexdataiq
Partner - Creator III
Partner - Creator III

I guess the

"After loading the document in memory (no cache) various calculations and aggregations wil create subsets of data that will be cached for real."

part that Peter was talking about meant that when a user make selections those calculations and aggregations will be cached.


Cheers

jjordaan
Partner - Specialist
Partner - Specialist

You can use the J-Meter for cache warming.

Quick tips #15 - Pre-caching with JMeter

Not applicable
Author

just read this


Objects that are on other sheets, minimized or hidden, are however not calculated.  from Logical Inference and Aggregations

Not applicable
Author

is there an example of creating a test script?

Peter_Cammaert
Partner - Champion III
Partner - Champion III

Pre-caching (filling the QVS cache with aggregations and expression results to improve selection and recalculation speed) is tricky stuff. I know of two techniques that used to be valid a while ago (not sure anymore). Stefan may have pointed to the first one already via an on-line demo by Miguel Garcia:

Best,

Peter

oknotsen
Master III
Master III

We (Qlik Education Services) cover this during our QlikView Performance and Scalability 11.2 training.

In short:

Look into J-Meter (Apache), QlikView ScriptGeneratorGIU and QlikView Results Analyzer, though you effectively only need the first two.

May you live in interesting times!
prees959
Creator II
Creator II

Hi, has anyone managed to get this working?  Or is there a more recent version of this (or an alternative) I can use to pre-warm a very large application?

Thanks All

Miguel_Angel_Baeyens

I have used the Qlik Scalability Tools, generating a script with the most likely actions (for example, activate each sheet) and a list of the users who open the app more often.

You need to schedule this out of the QMC, but the Windows scheduler will work just fine, and since it is JMeter based and can be used via command line, you can also integrate with other tools.

See here: Quick tips #15 - Pre-caching with JMeter‌. There might be a more recent version of the document, which I have not handy.

prees959
Creator II
Creator II

Hi MIguel,

Thanks for your reply.  I have installed Jmeter 5 and QlikScalabilitytools and I have run the following script :

jmeter_Testing.PNG

Can you suggest what to do next as the intructions are a bit vague?

Many thanks for your help,,

Phil

Miguel_Angel_Baeyens

In that document, you have a command line script you have to execute after the reload of the new data and prior to the access by the users. Using Windows Task Scheduler will do.

What you need to do is to define is the click script which will better serve the purpose of cache warming given that:

  1. Not all possible combinations are possible (it would take too long)
  2. There are some VIP users which will have priority

For scenario 1 create this click script using the script generator GUI included in the Scalability Tools. Actions such as Activate all Sheets, make "default" selections if there are any (e.g.: select current month), activate heavier charts or make selections on more populated sheets will always help.

For the second, create a list of users as if they were in the section access (if there is already section access in the document you will need that anyway). The idea is that during the time it takes to perform the script, the server is going to be utilized very, very heavily, so the entire script has to be short (depending on your users and reloads, I would say no longer than 15-20 minutes).

I'm still getting accustomed to the new Community, bear with while I retrieve links I had bookmarked for the previous.