1 Reply Latest reply: Apr 27, 2016 5:45 PM by Stefan Wühl RSS

    IRR - Slow Performance

    John Watson

      Has anyone experienced performance issues with QlikView's IRR function?

       

      Here's the scenario - we have a star schema data model with 2 fact tables concatenated. We've already done everything we could to improve performance but our application was still slow in AccessPoint compared to other comparable applications, at around 3 to 8 seconds delay when filtering. The QVW is only ~80MB, RAM usage is low, and the CPU usage is observed as being high. After some fiddling we found that it was our IRR formula that was causing the significant delay - when removed the other calculations performed fairly quickly. We tried getting the load order of the cash flows chronologically by date, but that did not help at all.

       

      Does anyone have any experience with the IRR function and possible recommendations for performance? I know that we can calculate the IRR in the script, but then we lose the ability to filter dynamically.

       

      Thanks.

        • Re: IRR - Slow Performance
          Stefan Wühl

          IRR calculates the internal rate of return using an iterative approximation, hence it might involve quite a lot of calculations / steps.

          The Excel function allows for a 'guess' parameter that might help to limit the number of iterations needed. There is no such parameter in QV.

           

          Not sure if your data model can be improved, but a simple star model with a single (concatenated) fact table looks quite good.

           

          Maybe the only possible improvement is to prevent QV from doing this calculations when it is not needed, e.g. using calculation conditions in your chart. Or somehow limit the scope for the calculations.