Unlock a world of possibilities! Login now and discover the exclusive benefits awaiting you.
Hello.
My old QlikView server have:
Intel(R)Xeon(R) 3.4 GHz (2 processor)
memory(RAM): 88 GB
My new qlikview server have:
AMD Opteron(tm) Processor 6380 2.50GHz(4 processors)
memory(RAM): 256 GB
But QlikView work very bad, old server is performance best.
I disabled numa nodes and virtual processors in bios disabled
This helped me, but performance server new = performance old server .
Help me configure my server.
Best regards
Yevgeniy
Hi,
Performance doesnt only depend on the Hardware configuration or on RAM, it also depends on how well the applications are designed and data models for that Applications are designed.
Regards,
Kaushik Solanki
Applications i do not change.
I change server
Have you use the RAM configuration, like it is in the mainboard manual for best performence. Can you disable error check of calculations in cpu. Have you used fast RAM or all bios options to speed it up.
Check if server use all cpu cores to calulate, how good is the performence in the client? Look for Bios updates.
Make benchmarks for RAM, HDD and CPU and compare them with the old Server to find the bottle neck.
Are all CPU Commant Sets activated, show if there is one to speed up calculations or slow it down.
I change all parameters on maximum performance.
I think, problem in AMD processor. I dont have more variants.
NUMA is a parameter for maximum performance, but creates conflicts with performances optimization from QV. This way other CPU speed up configurations can destroy QV optimization to, there is no other way as test different settings.
As customer or partner you can contact official support to get help from the QlikTech scalability labs in Lund.
I writing on top
I disabled numa nodes and virtual processors in bios disabled
This helped me, but performance server new = performance old server .
The main point i want to make is, if you can set a parameter on max performance it speed up only special actions. But it mostly have a secondary effect on other actions. Most knowledge of the cpu usage from QV, research, test and shared daily experience from partners are in the QT labs in Sweden.