QlikView Server CPU use

[Senor_Dai]

HI. we have a 32 Core VM running our live qlikview server.

We are planning a migration to QlikSense Enterprise or QlikSense Analytics Platform.

When I check theQV Governance dashboard I notice that the Max CPU usage is below 50% .. Does this suggest that we could use 16 cores with the same performance on our current QV instance?  And also would this translate to 16 cores in QAP?

Many thanks.

Dai

[Maria_Halley]

@Senor_Dai

You could probably use less cores, but I would not go down to 16 cores. If you do that your CPU usage would probably be very high.

Since it is a VM I would suggest that you try lowering it in steps and measure performance. 

Since QlikView and QlikSense are different products, so I can't say if 16 cores for QAP is suitable.

[marcus_sommer · Accepted solution]

This view might not show the whole truth because it depends in which frequencies the data are logged. AFAIK it are by default each 5 minutes. This means you see just one value for 5 minutes and the CPU peaks in between may be much higher. Beside this I'm not sure if Qlik will include all work-load scenarios in regard to double processor systems, virtual cores and similar stuff.

Personally I would expect to see always 100% usage within the max. A quite simple check would be to watch a bit the task-manager of the server during a busy time.

- Marcus

[Senor_Dai]

Thanks @Maria_Halley  and @marcus_sommer ! Much appreciated

[Maria_Halley]

@marcus_sommer

Can you explain that you mean by "I would expect to see always 100% usage within the max?

[marcus_sommer]

I did assume that this called max. value - AFAIK based on the field "CPULoad" - contained the max. value of the CPU workload. But depending on your question I'm not sure that this interpretation is correct - it may be just an average value. I would appreciate to get here more information.

But nevertheless I think the above given suggestion that this value couldn't show the whole truth is further correct and in between the 5 minutes logging interval could the max. CPU workload be much higher and if the datamodels and the UI are well-designed the calculations should be scaling over all cores and taking all available resources and therefore hitting regularly the 100% - ideally always rather short peaks.

If I'm now right that's an average this value displayed the mid-term workload and for this is a 5 minute interval a good default-setting and quite suitable to hint for an overload of the CPU. 

- Marcus