Unlock a world of possibilities! Login now and discover the exclusive benefits awaiting you.
Hi everyone
To increase our Qlik Sense Server's Performance, I thought it was a good idea to increase CPU cores and RAM.
Also in the Operations Monitor (see image below), our CPU usage was at a Maximum this August with 90%.
Our RAM usage also often has been at 80% + and from time to time our Server crashed. So I asked our IT to increase CPU and RAM, but they told me it wouldn't help anything: Their measures say that max CPU usage in August was at 38% and max RAM usage at 86%.
So my questions are:
What Kind of usage is the max CPU usage in Operations Monitor? Shouldn't it be the same number as the one our IT delivered?
And whatever number it is: Can't more CPU cores increase Qlik Sense's Performance anyway?
And what about RAM?
I'm really no Computer Hardware specialist and hope you can help me with That less Information given.
Best regards
Michael
Quite probably could you get near the limits of your environment and/or maybe touch it sometimes by round 5 GB on qvd-data and only 16 GB RAM.
I think the statements from your IT aren't quite correct than many load-statements and gui-calculations are fully multi-threaded so that AFAIK it must be 100% of CPU consumption quite often. A RAM consumption of 86% (if it's really the truth) is a bad sign than it meant that there nearly no reserves left and at least the OS will swap their RAM heavily within the virtual RAM on the HD.
If this swapping happens with QLIK it will be very slow and what you described fits in this pattern. I'm not sure how qlik sense certain server settings handled but I think it will like qlikview have some working set settings - which control the min/max values of used RAM and further configurations how many cores are in use and some more.
Another point which might help (and which is often possible) is to optimize the datamodel and the gui.
- Marcus
Hi Michael,
Can you share the CPU and RAM numbers please.
Regards,
Karthik
Hi Karthikeyan
Sure, RAM: 16GB, CPU: "Intel(R) Xeon(R) CPU-E52650 v3 @2.30GHz (4 CPUs)" whatever this means 😉
Regards
Michael
Hi Mike,
16 GB is very low. Need Number of Cores information also can you share the OS information please.
Hi Michael,
Can you please tell me , what is the size of data ,second what kind of performance you are talking about, like loading application on browser and all ?
Hi Karthikeyan
OS is Windows Server 2012 R2
Number of Cores:
In Task Manager it says 4 sockets and 4 virtual cores.
thx
Hi Rohit
In all QVDs together we're storing about 5GB and we evaluate everything directly in this base.
Performance is just everywhere bad: It takes a few seconds to open apps, it takes several minutes to load data from QVDs if there are many line operations/transformations in the load script etc.
So my question is basically if you think more RAM/CPU cores could help.
Thanks!
Quite probably could you get near the limits of your environment and/or maybe touch it sometimes by round 5 GB on qvd-data and only 16 GB RAM.
I think the statements from your IT aren't quite correct than many load-statements and gui-calculations are fully multi-threaded so that AFAIK it must be 100% of CPU consumption quite often. A RAM consumption of 86% (if it's really the truth) is a bad sign than it meant that there nearly no reserves left and at least the OS will swap their RAM heavily within the virtual RAM on the HD.
If this swapping happens with QLIK it will be very slow and what you described fits in this pattern. I'm not sure how qlik sense certain server settings handled but I think it will like qlikview have some working set settings - which control the min/max values of used RAM and further configurations how many cores are in use and some more.
Another point which might help (and which is often possible) is to optimize the datamodel and the gui.
- Marcus
Thanks for your Response, Marcus!
Will have a more detailled look at it together with our IT.
Regards
Michael
Hi Michael,
Few points below which can help you to improve performance :
1. 3-tier Architecture
2. data model with best practices.
3. Try to avoid joins as much as you can , try to make a flat file at DB then connect to it and create QVD with every table.
4. I don't think for 5 GB data 16 GB is less.