Assessing performance and optimisation - problems with calculation time

[#1]

Hi All,

I'm in the process of trying to optimise a document.  It has lots of data in it and performance is not that great.  So basically I am trying to implement a bunch of ideas from Qlik and the community.

What I'm having trouble with is assessing whether they actually result in improvements at an object level.  The only way I know how to do this is to open the document fresh, select the page I want to assess, make a couple of selections, then export the memory statistics.  Then I can compare before v after.

The problems are:

- This is a very slow and inefficient process, is there a tool that can be used instead?

- The results are extremely inconsistent.  Same data, same object can give me radically different calculation times.  Also, the new v optimised can show me that one is 'faster' than the other one time, but then the opposite another time.

Is there a better approach?  I'm starting to lose all faith in the calculation times for assessment.

Cheers,

GPC

[Gysbert_Wassenaar]

See this blog post: Recipe for a Memory Statistics analysis

[#3]

Hi Gysbert,

Thanks for the reply - that is a useful blog post.

I'm more concerned with knowing if a change has resulted in an improvement.  When that improvement is small it seems that the memory statistics are pretty unreliable.

Cheers,

G

[shane_spencer]

"- The results are extremely inconsistent.  Same data, same object can give me radically different calculation times.  Also, the new v optimised can show me that one is 'faster' than the other one time, but then the opposite another time."

Have you restarted QVS between each test to ensure that you have no cached results? Also have you made sure nothing else is running on the server?