Performance When Increasing Number of Processor Cores

[#1]

Given the same amount of ram, is Qlikview faster or more efficient if you increase processor power by adding cores? If so, is their a limit to the amount of cores Qlikview can use?

Thanks for the Help,

Stephen

[pover]

QlikView is not necessarily faster with more processor cores, but it should handle more users as the number of processor cores increases. There is no limit as far as I know. The biggest system we have has 32 processing cores. We are still doing tests, but review this post to avoid problems when you increase of the number of cores.

Regards.

[danielrozental]

By "this post" Karl probably meant this post http://community.qlik.com/forums/p/33509/139595.aspx

[pover]

Thank you.

[#5]

Hi,

QlikView scales very nicely by adding "real" CPU cores. It means faster for few users, and faster for many users. Having several servers with 24 and 32 cores.

But be aware, performance is really bad if CPU hyper-threading is enabled = the default setting

-Alex

[pover]

Alex is right although we've seen that with CPUs with less cores, hyperthreading can help and there are a variety of other configurations that help or hurt performance, but the best configuration doesn't appear to be universal. We've been comparing a variety of servers with 32 physical and logical cores with mixed results even though I mention just one particular server in the post.

There is also the topic of clustering servers to increase stability and performance that you might want to investigate more.

After my experience the only thing I can recommend is that you be careful when selling the idea of increased performance with more CPUs. It doesn't necessarily mean that QlikView is going to be faster, but it should mean that you can support more users. Additionally, be prepared to adjust the server configuration to get the best results.

Good luck.

[dirk_konings]

as of QVS 10, QVS will use more CPU to do reloads and memory calculations. This was not before QV10. Then QVS used a lot of memory and a small part of cpu.

[pover]

The reload process in QlikView 10 uses more CPUs and is noticebly faster, but the chart calculations appear to take the same amount of time or are even a little slower than the chart calculations in QV 9. Also, QV 9 uses all the CPUs possible to calculate charts.

Maybe some expressions, for example calculated dimensions, calculate faster in QV 10, but I've only tested well formulated, simple charts and I have not seen any improvement in their calculation time.

Regards.