Performance of documents on the server vs desktop

[#1]

Hi,

I'm after other QV developers experience of the difference in performance between o document on the desktop and the same document, using Ajax on the Qlikview server. My experience is that a document that works very quickly, with almost instant response to clicks and filters slows down significantly on the server through IE8 web browser. Instant reposes become 3 to 4 seconds, 1 second responses become 10-11 seconds. The feedback from users is that it is slower than they expected as we tend to demo using a local installation.

Our server is a few years old and we have a new one in the pipeline, but even on small documents (20,000 records, 2MB file, efficient expressions) we see a signficant  loss of performance.

Anyone had a similar experience, does anyone have any advice, will a brand new server with more RAM, faster processes help?

Appreciate your thoughts.

Matt

[erichshiino]

Hi,

Did you get a different performance only using Ajax? Did you try IE Plugin?Did you try different browsers?

I know there is a server issue with hyperthreading. For more information, go to this thread:

http://community.qlik.com/message/79581#79581

Regards,

Erich

[#3]

Hi Erich,

Have tried different browsers, and the IE plugin. Plugin is slightly better, but not by much - we really need to use the Ajax zero footprint / IE browser in our organsiation due to the desktop policy that we have.

Will take a look into the hyperthreading, thank you.

One thing I failed to mention as well is that the server we are currently using is in the states,  bulk of the users (and myself) are UK based - we're all on the same WAN, but coiuld the delays be down to latency on the network? I'm gathering response times from the states to see how they differ.

Matt

[csellei]

Hi,

Are you working with a virtual server?

I have a similar issue and the reason was virtualization.

[danielrozental]

We also had a performance issue, where CPU usage would go to 100% for several seconds, with Ajax and QV10 that was resolved with SR3.